xref: /freebsd/sys/net/if_bridge.c (revision fba91af3b09b0cb021a50da2bc78e44dfd49b69a)
1 /*	$NetBSD: if_bridge.c,v 1.31 2005/06/01 19:45:34 jdc Exp $	*/
2 
3 /*-
4  * SPDX-License-Identifier: BSD-4-Clause
5  *
6  * Copyright 2001 Wasabi Systems, Inc.
7  * All rights reserved.
8  *
9  * Written by Jason R. Thorpe for Wasabi Systems, Inc.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed for the NetBSD Project by
22  *	Wasabi Systems, Inc.
23  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
24  *    or promote products derived from this software without specific prior
25  *    written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net)
42  * All rights reserved.
43  *
44  * Redistribution and use in source and binary forms, with or without
45  * modification, are permitted provided that the following conditions
46  * are met:
47  * 1. Redistributions of source code must retain the above copyright
48  *    notice, this list of conditions and the following disclaimer.
49  * 2. Redistributions in binary form must reproduce the above copyright
50  *    notice, this list of conditions and the following disclaimer in the
51  *    documentation and/or other materials provided with the distribution.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
54  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
55  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
56  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
57  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
58  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
59  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
61  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
62  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
63  * POSSIBILITY OF SUCH DAMAGE.
64  *
65  * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp
66  */
67 
68 /*
69  * Network interface bridge support.
70  *
71  * TODO:
72  *
73  *	- Currently only supports Ethernet-like interfaces (Ethernet,
74  *	  802.11, VLANs on Ethernet, etc.)  Figure out a nice way
75  *	  to bridge other types of interfaces (maybe consider
76  *	  heterogeneous bridges).
77  */
78 
79 #include <sys/cdefs.h>
80 #include "opt_inet.h"
81 #include "opt_inet6.h"
82 
83 #include <sys/param.h>
84 #include <sys/eventhandler.h>
85 #include <sys/mbuf.h>
86 #include <sys/malloc.h>
87 #include <sys/protosw.h>
88 #include <sys/systm.h>
89 #include <sys/jail.h>
90 #include <sys/time.h>
91 #include <sys/socket.h> /* for net/if.h */
92 #include <sys/sockio.h>
93 #include <sys/ctype.h>  /* string functions */
94 #include <sys/kernel.h>
95 #include <sys/random.h>
96 #include <sys/syslog.h>
97 #include <sys/sysctl.h>
98 #include <vm/uma.h>
99 #include <sys/module.h>
100 #include <sys/priv.h>
101 #include <sys/proc.h>
102 #include <sys/lock.h>
103 #include <sys/mutex.h>
104 
105 #include <net/bpf.h>
106 #include <net/if.h>
107 #include <net/if_clone.h>
108 #include <net/if_dl.h>
109 #include <net/if_types.h>
110 #include <net/if_var.h>
111 #include <net/if_private.h>
112 #include <net/pfil.h>
113 #include <net/vnet.h>
114 
115 #include <netinet/in.h>
116 #include <netinet/in_systm.h>
117 #include <netinet/in_var.h>
118 #include <netinet/ip.h>
119 #include <netinet/ip_var.h>
120 #ifdef INET6
121 #include <netinet/ip6.h>
122 #include <netinet6/ip6_var.h>
123 #include <netinet6/in6_ifattach.h>
124 #endif
125 #if defined(INET) || defined(INET6)
126 #include <netinet/ip_carp.h>
127 #endif
128 #include <machine/in_cksum.h>
129 #include <netinet/if_ether.h>
130 #include <net/bridgestp.h>
131 #include <net/if_bridgevar.h>
132 #include <net/if_llc.h>
133 #include <net/if_vlan_var.h>
134 
135 #include <net/route.h>
136 
137 /*
138  * At various points in the code we need to know if we're hooked into the INET
139  * and/or INET6 pfil.  Define some macros to do that based on which IP versions
140  * are enabled in the kernel.  This avoids littering the rest of the code with
141  * #ifnet INET6 to avoid referencing V_inet6_pfil_head.
142  */
143 #ifdef INET6
144 #define		PFIL_HOOKED_IN_INET6	PFIL_HOOKED_IN(V_inet6_pfil_head)
145 #define		PFIL_HOOKED_OUT_INET6	PFIL_HOOKED_OUT(V_inet6_pfil_head)
146 #else
147 #define		PFIL_HOOKED_IN_INET6	false
148 #define		PFIL_HOOKED_OUT_INET6	false
149 #endif
150 
151 #ifdef INET
152 #define		PFIL_HOOKED_IN_INET	PFIL_HOOKED_IN(V_inet_pfil_head)
153 #define		PFIL_HOOKED_OUT_INET	PFIL_HOOKED_OUT(V_inet_pfil_head)
154 #else
155 #define		PFIL_HOOKED_IN_INET	false
156 #define		PFIL_HOOKED_OUT_INET	false
157 #endif
158 
159 #define		PFIL_HOOKED_IN_46	(PFIL_HOOKED_IN_INET6 || PFIL_HOOKED_IN_INET)
160 #define		PFIL_HOOKED_OUT_46	(PFIL_HOOKED_OUT_INET6 || PFIL_HOOKED_OUT_INET)
161 
162 /*
163  * Size of the route hash table.  Must be a power of two.
164  */
165 #ifndef BRIDGE_RTHASH_SIZE
166 #define	BRIDGE_RTHASH_SIZE		1024
167 #endif
168 
169 #define	BRIDGE_RTHASH_MASK		(BRIDGE_RTHASH_SIZE - 1)
170 
171 /*
172  * Default maximum number of addresses to cache.
173  */
174 #ifndef BRIDGE_RTABLE_MAX
175 #define	BRIDGE_RTABLE_MAX		2000
176 #endif
177 
178 /*
179  * Timeout (in seconds) for entries learned dynamically.
180  */
181 #ifndef BRIDGE_RTABLE_TIMEOUT
182 #define	BRIDGE_RTABLE_TIMEOUT		(20 * 60)	/* same as ARP */
183 #endif
184 
185 /*
186  * Number of seconds between walks of the route list.
187  */
188 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
189 #define	BRIDGE_RTABLE_PRUNE_PERIOD	(5 * 60)
190 #endif
191 
192 /*
193  * List of capabilities to possibly mask on the member interface.
194  */
195 #define	BRIDGE_IFCAPS_MASK		(IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM|\
196 					 IFCAP_TXCSUM_IPV6|IFCAP_MEXTPG)
197 
198 /*
199  * List of capabilities to strip
200  */
201 #define	BRIDGE_IFCAPS_STRIP		IFCAP_LRO
202 
203 /*
204  * Bridge locking
205  *
206  * The bridge relies heavily on the epoch(9) system to protect its data
207  * structures. This means we can safely use CK_LISTs while in NET_EPOCH, but we
208  * must ensure there is only one writer at a time.
209  *
210  * That is: for read accesses we only need to be in NET_EPOCH, but for write
211  * accesses we must hold:
212  *
213  *  - BRIDGE_RT_LOCK, for any change to bridge_rtnodes
214  *  - BRIDGE_LOCK, for any other change
215  *
216  * The BRIDGE_LOCK is a sleepable lock, because it is held across ioctl()
217  * calls to bridge member interfaces and these ioctl()s can sleep.
218  * The BRIDGE_RT_LOCK is a non-sleepable mutex, because it is sometimes
219  * required while we're in NET_EPOCH and then we're not allowed to sleep.
220  */
221 #define BRIDGE_LOCK_INIT(_sc)		do {			\
222 	sx_init(&(_sc)->sc_sx, "if_bridge");			\
223 	mtx_init(&(_sc)->sc_rt_mtx, "if_bridge rt", NULL, MTX_DEF);	\
224 } while (0)
225 #define BRIDGE_LOCK_DESTROY(_sc)	do {	\
226 	sx_destroy(&(_sc)->sc_sx);		\
227 	mtx_destroy(&(_sc)->sc_rt_mtx);		\
228 } while (0)
229 #define BRIDGE_LOCK(_sc)		sx_xlock(&(_sc)->sc_sx)
230 #define BRIDGE_UNLOCK(_sc)		sx_xunlock(&(_sc)->sc_sx)
231 #define BRIDGE_LOCK_ASSERT(_sc)		sx_assert(&(_sc)->sc_sx, SX_XLOCKED)
232 #define BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(_sc)	\
233 	    MPASS(in_epoch(net_epoch_preempt) || sx_xlocked(&(_sc)->sc_sx))
234 #define BRIDGE_UNLOCK_ASSERT(_sc)	sx_assert(&(_sc)->sc_sx, SX_UNLOCKED)
235 #define BRIDGE_RT_LOCK(_sc)		mtx_lock(&(_sc)->sc_rt_mtx)
236 #define BRIDGE_RT_UNLOCK(_sc)		mtx_unlock(&(_sc)->sc_rt_mtx)
237 #define BRIDGE_RT_LOCK_ASSERT(_sc)	mtx_assert(&(_sc)->sc_rt_mtx, MA_OWNED)
238 #define BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(_sc)	\
239 	    MPASS(in_epoch(net_epoch_preempt) || mtx_owned(&(_sc)->sc_rt_mtx))
240 
241 struct bridge_softc;
242 
243 /*
244  * Bridge interface list entry.
245  */
246 struct bridge_iflist {
247 	CK_LIST_ENTRY(bridge_iflist) bif_next;
248 	struct ifnet		*bif_ifp;	/* member if */
249 	struct bridge_softc	*bif_sc;	/* parent bridge */
250 	struct bstp_port	bif_stp;	/* STP state */
251 	uint32_t		bif_flags;	/* member if flags */
252 	int			bif_savedcaps;	/* saved capabilities */
253 	uint32_t		bif_addrmax;	/* max # of addresses */
254 	uint32_t		bif_addrcnt;	/* cur. # of addresses */
255 	uint32_t		bif_addrexceeded;/* # of address violations */
256 	struct epoch_context	bif_epoch_ctx;
257 };
258 
259 /*
260  * Bridge route node.
261  */
262 struct bridge_rtnode {
263 	CK_LIST_ENTRY(bridge_rtnode) brt_hash;	/* hash table linkage */
264 	CK_LIST_ENTRY(bridge_rtnode) brt_list;	/* list linkage */
265 	struct bridge_iflist	*brt_dst;	/* destination if */
266 	unsigned long		brt_expire;	/* expiration time */
267 	uint8_t			brt_flags;	/* address flags */
268 	uint8_t			brt_addr[ETHER_ADDR_LEN];
269 	ether_vlanid_t		brt_vlan;	/* vlan id */
270 	struct	vnet		*brt_vnet;
271 	struct	epoch_context	brt_epoch_ctx;
272 };
273 #define	brt_ifp			brt_dst->bif_ifp
274 
275 /*
276  * Software state for each bridge.
277  */
278 struct bridge_softc {
279 	struct ifnet		*sc_ifp;	/* make this an interface */
280 	LIST_ENTRY(bridge_softc) sc_list;
281 	struct sx		sc_sx;
282 	struct mtx		sc_rt_mtx;
283 	uint32_t		sc_brtmax;	/* max # of addresses */
284 	uint32_t		sc_brtcnt;	/* cur. # of addresses */
285 	uint32_t		sc_brttimeout;	/* rt timeout in seconds */
286 	struct callout		sc_brcallout;	/* bridge callout */
287 	CK_LIST_HEAD(, bridge_iflist) sc_iflist;	/* member interface list */
288 	CK_LIST_HEAD(, bridge_rtnode) *sc_rthash;	/* our forwarding table */
289 	CK_LIST_HEAD(, bridge_rtnode) sc_rtlist;	/* list version of above */
290 	uint32_t		sc_rthash_key;	/* key for hash */
291 	CK_LIST_HEAD(, bridge_iflist) sc_spanlist;	/* span ports list */
292 	struct bstp_state	sc_stp;		/* STP state */
293 	uint32_t		sc_brtexceeded;	/* # of cache drops */
294 	struct ifnet		*sc_ifaddr;	/* member mac copied from */
295 	struct ether_addr	sc_defaddr;	/* Default MAC address */
296 	if_input_fn_t		sc_if_input;	/* Saved copy of if_input */
297 	struct epoch_context	sc_epoch_ctx;
298 };
299 
300 VNET_DEFINE_STATIC(struct sx, bridge_list_sx);
301 #define	V_bridge_list_sx	VNET(bridge_list_sx)
302 static eventhandler_tag bridge_detach_cookie;
303 
304 int	bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
305 
306 VNET_DEFINE_STATIC(uma_zone_t, bridge_rtnode_zone);
307 #define	V_bridge_rtnode_zone	VNET(bridge_rtnode_zone)
308 
309 static int	bridge_clone_create(struct if_clone *, char *, size_t,
310 		    struct ifc_data *, struct ifnet **);
311 static int	bridge_clone_destroy(struct if_clone *, struct ifnet *, uint32_t);
312 
313 static int	bridge_ioctl(struct ifnet *, u_long, caddr_t);
314 static void	bridge_mutecaps(struct bridge_softc *);
315 static void	bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
316 		    int);
317 static void	bridge_ifdetach(void *arg __unused, struct ifnet *);
318 static void	bridge_init(void *);
319 static void	bridge_dummynet(struct mbuf *, struct ifnet *);
320 static bool	bridge_same(const void *, const void *);
321 static void	*bridge_get_softc(struct ifnet *);
322 static void	bridge_stop(struct ifnet *, int);
323 static int	bridge_transmit(struct ifnet *, struct mbuf *);
324 #ifdef ALTQ
325 static void	bridge_altq_start(if_t);
326 static int	bridge_altq_transmit(if_t, struct mbuf *);
327 #endif
328 static void	bridge_qflush(struct ifnet *);
329 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
330 static void	bridge_inject(struct ifnet *, struct mbuf *);
331 static int	bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
332 		    struct rtentry *);
333 static int	bridge_enqueue(struct bridge_softc *, struct ifnet *,
334 		    struct mbuf *);
335 static void	bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
336 
337 static void	bridge_forward(struct bridge_softc *, struct bridge_iflist *,
338 		    struct mbuf *m);
339 static bool	bridge_member_ifaddrs(void);
340 
341 static void	bridge_timer(void *);
342 
343 static void	bridge_broadcast(struct bridge_softc *, struct ifnet *,
344 		    struct mbuf *, int);
345 static void	bridge_span(struct bridge_softc *, struct mbuf *);
346 
347 static int	bridge_rtupdate(struct bridge_softc *, const uint8_t *,
348 		    ether_vlanid_t, struct bridge_iflist *, int, uint8_t);
349 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
350 		    ether_vlanid_t);
351 static void	bridge_rttrim(struct bridge_softc *);
352 static void	bridge_rtage(struct bridge_softc *);
353 static void	bridge_rtflush(struct bridge_softc *, int);
354 static int	bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
355 		    ether_vlanid_t);
356 
357 static void	bridge_rtable_init(struct bridge_softc *);
358 static void	bridge_rtable_fini(struct bridge_softc *);
359 
360 static int	bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
361 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
362 		    const uint8_t *, ether_vlanid_t);
363 static int	bridge_rtnode_insert(struct bridge_softc *,
364 		    struct bridge_rtnode *);
365 static void	bridge_rtnode_destroy(struct bridge_softc *,
366 		    struct bridge_rtnode *);
367 static void	bridge_rtable_expire(struct ifnet *, int);
368 static void	bridge_state_change(struct ifnet *, int);
369 
370 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
371 		    const char *name);
372 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
373 		    struct ifnet *ifp);
374 static void	bridge_delete_member(struct bridge_softc *,
375 		    struct bridge_iflist *, int);
376 static void	bridge_delete_span(struct bridge_softc *,
377 		    struct bridge_iflist *);
378 
379 static int	bridge_ioctl_add(struct bridge_softc *, void *);
380 static int	bridge_ioctl_del(struct bridge_softc *, void *);
381 static int	bridge_ioctl_gifflags(struct bridge_softc *, void *);
382 static int	bridge_ioctl_sifflags(struct bridge_softc *, void *);
383 static int	bridge_ioctl_scache(struct bridge_softc *, void *);
384 static int	bridge_ioctl_gcache(struct bridge_softc *, void *);
385 static int	bridge_ioctl_gifs(struct bridge_softc *, void *);
386 static int	bridge_ioctl_rts(struct bridge_softc *, void *);
387 static int	bridge_ioctl_saddr(struct bridge_softc *, void *);
388 static int	bridge_ioctl_sto(struct bridge_softc *, void *);
389 static int	bridge_ioctl_gto(struct bridge_softc *, void *);
390 static int	bridge_ioctl_daddr(struct bridge_softc *, void *);
391 static int	bridge_ioctl_flush(struct bridge_softc *, void *);
392 static int	bridge_ioctl_gpri(struct bridge_softc *, void *);
393 static int	bridge_ioctl_spri(struct bridge_softc *, void *);
394 static int	bridge_ioctl_ght(struct bridge_softc *, void *);
395 static int	bridge_ioctl_sht(struct bridge_softc *, void *);
396 static int	bridge_ioctl_gfd(struct bridge_softc *, void *);
397 static int	bridge_ioctl_sfd(struct bridge_softc *, void *);
398 static int	bridge_ioctl_gma(struct bridge_softc *, void *);
399 static int	bridge_ioctl_sma(struct bridge_softc *, void *);
400 static int	bridge_ioctl_sifprio(struct bridge_softc *, void *);
401 static int	bridge_ioctl_sifcost(struct bridge_softc *, void *);
402 static int	bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
403 static int	bridge_ioctl_addspan(struct bridge_softc *, void *);
404 static int	bridge_ioctl_delspan(struct bridge_softc *, void *);
405 static int	bridge_ioctl_gbparam(struct bridge_softc *, void *);
406 static int	bridge_ioctl_grte(struct bridge_softc *, void *);
407 static int	bridge_ioctl_gifsstp(struct bridge_softc *, void *);
408 static int	bridge_ioctl_sproto(struct bridge_softc *, void *);
409 static int	bridge_ioctl_stxhc(struct bridge_softc *, void *);
410 static int	bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
411 		    int);
412 #ifdef INET
413 static int	bridge_ip_checkbasic(struct mbuf **mp);
414 static int	bridge_fragment(struct ifnet *, struct mbuf **mp,
415 		    struct ether_header *, int, struct llc *);
416 #endif /* INET */
417 #ifdef INET6
418 static int	bridge_ip6_checkbasic(struct mbuf **mp);
419 #endif /* INET6 */
420 static void	bridge_linkstate(struct ifnet *ifp);
421 static void	bridge_linkcheck(struct bridge_softc *sc);
422 
423 /*
424  * Use the "null" value from IEEE 802.1Q-2014 Table 9-2
425  * to indicate untagged frames.
426  */
427 #define	VLANTAGOF(_m)	\
428     ((_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : DOT1Q_VID_NULL)
429 
430 static struct bstp_cb_ops bridge_ops = {
431 	.bcb_state = bridge_state_change,
432 	.bcb_rtage = bridge_rtable_expire
433 };
434 
435 SYSCTL_DECL(_net_link);
436 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
437     "Bridge");
438 
439 /* only pass IP[46] packets when pfil is enabled */
440 VNET_DEFINE_STATIC(int, pfil_onlyip) = 1;
441 #define	V_pfil_onlyip	VNET(pfil_onlyip)
442 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip,
443     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_onlyip), 0,
444     "Only pass IP packets when pfil is enabled");
445 
446 /* run pfil hooks on the bridge interface */
447 VNET_DEFINE_STATIC(int, pfil_bridge) = 0;
448 #define	V_pfil_bridge	VNET(pfil_bridge)
449 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge,
450     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_bridge), 0,
451     "Packet filter on the bridge interface");
452 
453 /* layer2 filter with ipfw */
454 VNET_DEFINE_STATIC(int, pfil_ipfw);
455 #define	V_pfil_ipfw	VNET(pfil_ipfw)
456 
457 /* layer2 ARP filter with ipfw */
458 VNET_DEFINE_STATIC(int, pfil_ipfw_arp);
459 #define	V_pfil_ipfw_arp	VNET(pfil_ipfw_arp)
460 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp,
461     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_ipfw_arp), 0,
462     "Filter ARP packets through IPFW layer2");
463 
464 /* run pfil hooks on the member interface */
465 VNET_DEFINE_STATIC(int, pfil_member) = 0;
466 #define	V_pfil_member	VNET(pfil_member)
467 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member,
468     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_member), 0,
469     "Packet filter on the member interface");
470 
471 /* run pfil hooks on the physical interface for locally destined packets */
472 VNET_DEFINE_STATIC(int, pfil_local_phys);
473 #define	V_pfil_local_phys	VNET(pfil_local_phys)
474 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys,
475     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_local_phys), 0,
476     "Packet filter on the physical interface for locally destined packets");
477 
478 /* log STP state changes */
479 VNET_DEFINE_STATIC(int, log_stp);
480 #define	V_log_stp	VNET(log_stp)
481 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp,
482     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(log_stp), 0,
483     "Log STP state changes");
484 
485 /* share MAC with first bridge member */
486 VNET_DEFINE_STATIC(int, bridge_inherit_mac);
487 #define	V_bridge_inherit_mac	VNET(bridge_inherit_mac)
488 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac,
489     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(bridge_inherit_mac), 0,
490     "Inherit MAC address from the first bridge member");
491 
492 VNET_DEFINE_STATIC(int, allow_llz_overlap) = 0;
493 #define	V_allow_llz_overlap	VNET(allow_llz_overlap)
494 SYSCTL_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap,
495     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(allow_llz_overlap), 0,
496     "Allow overlap of link-local scope "
497     "zones of a bridge interface and the member interfaces");
498 
499 /* log MAC address port flapping */
500 VNET_DEFINE_STATIC(bool, log_mac_flap) = true;
501 #define	V_log_mac_flap	VNET(log_mac_flap)
502 SYSCTL_BOOL(_net_link_bridge, OID_AUTO, log_mac_flap,
503     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(log_mac_flap), true,
504     "Log MAC address port flapping");
505 
506 /* allow IP addresses on bridge members */
507 VNET_DEFINE_STATIC(bool, member_ifaddrs) = false;
508 #define	V_member_ifaddrs	VNET(member_ifaddrs)
509 SYSCTL_BOOL(_net_link_bridge, OID_AUTO, member_ifaddrs,
510     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(member_ifaddrs), false,
511     "Allow layer 3 addresses on bridge members");
512 
513 static bool
514 bridge_member_ifaddrs(void)
515 {
516 	return (V_member_ifaddrs);
517 }
518 
519 VNET_DEFINE_STATIC(int, log_interval) = 5;
520 VNET_DEFINE_STATIC(int, log_count) = 0;
521 VNET_DEFINE_STATIC(struct timeval, log_last) = { 0 };
522 
523 #define	V_log_interval	VNET(log_interval)
524 #define	V_log_count	VNET(log_count)
525 #define	V_log_last	VNET(log_last)
526 
527 struct bridge_control {
528 	int	(*bc_func)(struct bridge_softc *, void *);
529 	int	bc_argsize;
530 	int	bc_flags;
531 };
532 
533 #define	BC_F_COPYIN		0x01	/* copy arguments in */
534 #define	BC_F_COPYOUT		0x02	/* copy arguments out */
535 #define	BC_F_SUSER		0x04	/* do super-user check */
536 
537 static const struct bridge_control bridge_control_table[] = {
538 	{ bridge_ioctl_add,		sizeof(struct ifbreq),
539 	  BC_F_COPYIN|BC_F_SUSER },
540 	{ bridge_ioctl_del,		sizeof(struct ifbreq),
541 	  BC_F_COPYIN|BC_F_SUSER },
542 
543 	{ bridge_ioctl_gifflags,	sizeof(struct ifbreq),
544 	  BC_F_COPYIN|BC_F_COPYOUT },
545 	{ bridge_ioctl_sifflags,	sizeof(struct ifbreq),
546 	  BC_F_COPYIN|BC_F_SUSER },
547 
548 	{ bridge_ioctl_scache,		sizeof(struct ifbrparam),
549 	  BC_F_COPYIN|BC_F_SUSER },
550 	{ bridge_ioctl_gcache,		sizeof(struct ifbrparam),
551 	  BC_F_COPYOUT },
552 
553 	{ bridge_ioctl_gifs,		sizeof(struct ifbifconf),
554 	  BC_F_COPYIN|BC_F_COPYOUT },
555 	{ bridge_ioctl_rts,		sizeof(struct ifbaconf),
556 	  BC_F_COPYIN|BC_F_COPYOUT },
557 
558 	{ bridge_ioctl_saddr,		sizeof(struct ifbareq),
559 	  BC_F_COPYIN|BC_F_SUSER },
560 
561 	{ bridge_ioctl_sto,		sizeof(struct ifbrparam),
562 	  BC_F_COPYIN|BC_F_SUSER },
563 	{ bridge_ioctl_gto,		sizeof(struct ifbrparam),
564 	  BC_F_COPYOUT },
565 
566 	{ bridge_ioctl_daddr,		sizeof(struct ifbareq),
567 	  BC_F_COPYIN|BC_F_SUSER },
568 
569 	{ bridge_ioctl_flush,		sizeof(struct ifbreq),
570 	  BC_F_COPYIN|BC_F_SUSER },
571 
572 	{ bridge_ioctl_gpri,		sizeof(struct ifbrparam),
573 	  BC_F_COPYOUT },
574 	{ bridge_ioctl_spri,		sizeof(struct ifbrparam),
575 	  BC_F_COPYIN|BC_F_SUSER },
576 
577 	{ bridge_ioctl_ght,		sizeof(struct ifbrparam),
578 	  BC_F_COPYOUT },
579 	{ bridge_ioctl_sht,		sizeof(struct ifbrparam),
580 	  BC_F_COPYIN|BC_F_SUSER },
581 
582 	{ bridge_ioctl_gfd,		sizeof(struct ifbrparam),
583 	  BC_F_COPYOUT },
584 	{ bridge_ioctl_sfd,		sizeof(struct ifbrparam),
585 	  BC_F_COPYIN|BC_F_SUSER },
586 
587 	{ bridge_ioctl_gma,		sizeof(struct ifbrparam),
588 	  BC_F_COPYOUT },
589 	{ bridge_ioctl_sma,		sizeof(struct ifbrparam),
590 	  BC_F_COPYIN|BC_F_SUSER },
591 
592 	{ bridge_ioctl_sifprio,		sizeof(struct ifbreq),
593 	  BC_F_COPYIN|BC_F_SUSER },
594 
595 	{ bridge_ioctl_sifcost,		sizeof(struct ifbreq),
596 	  BC_F_COPYIN|BC_F_SUSER },
597 
598 	{ bridge_ioctl_addspan,		sizeof(struct ifbreq),
599 	  BC_F_COPYIN|BC_F_SUSER },
600 	{ bridge_ioctl_delspan,		sizeof(struct ifbreq),
601 	  BC_F_COPYIN|BC_F_SUSER },
602 
603 	{ bridge_ioctl_gbparam,		sizeof(struct ifbropreq),
604 	  BC_F_COPYOUT },
605 
606 	{ bridge_ioctl_grte,		sizeof(struct ifbrparam),
607 	  BC_F_COPYOUT },
608 
609 	{ bridge_ioctl_gifsstp,		sizeof(struct ifbpstpconf),
610 	  BC_F_COPYIN|BC_F_COPYOUT },
611 
612 	{ bridge_ioctl_sproto,		sizeof(struct ifbrparam),
613 	  BC_F_COPYIN|BC_F_SUSER },
614 
615 	{ bridge_ioctl_stxhc,		sizeof(struct ifbrparam),
616 	  BC_F_COPYIN|BC_F_SUSER },
617 
618 	{ bridge_ioctl_sifmaxaddr,	sizeof(struct ifbreq),
619 	  BC_F_COPYIN|BC_F_SUSER },
620 
621 };
622 static const int bridge_control_table_size = nitems(bridge_control_table);
623 
624 VNET_DEFINE_STATIC(LIST_HEAD(, bridge_softc), bridge_list) =
625     LIST_HEAD_INITIALIZER();
626 #define	V_bridge_list	VNET(bridge_list)
627 #define	BRIDGE_LIST_LOCK_INIT(x)	sx_init(&V_bridge_list_sx,	\
628 					    "if_bridge list")
629 #define	BRIDGE_LIST_LOCK_DESTROY(x)	sx_destroy(&V_bridge_list_sx)
630 #define	BRIDGE_LIST_LOCK(x)		sx_xlock(&V_bridge_list_sx)
631 #define	BRIDGE_LIST_UNLOCK(x)		sx_xunlock(&V_bridge_list_sx)
632 
633 VNET_DEFINE_STATIC(struct if_clone *, bridge_cloner);
634 #define	V_bridge_cloner	VNET(bridge_cloner)
635 
636 static const char bridge_name[] = "bridge";
637 
638 static void
639 vnet_bridge_init(const void *unused __unused)
640 {
641 
642 	V_bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
643 	    sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
644 	    UMA_ALIGN_PTR, 0);
645 	BRIDGE_LIST_LOCK_INIT();
646 
647 	struct if_clone_addreq req = {
648 		.create_f = bridge_clone_create,
649 		.destroy_f = bridge_clone_destroy,
650 		.flags = IFC_F_AUTOUNIT,
651 	};
652 	V_bridge_cloner = ifc_attach_cloner(bridge_name, &req);
653 }
654 VNET_SYSINIT(vnet_bridge_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
655     vnet_bridge_init, NULL);
656 
657 static void
658 vnet_bridge_uninit(const void *unused __unused)
659 {
660 
661 	ifc_detach_cloner(V_bridge_cloner);
662 	V_bridge_cloner = NULL;
663 	BRIDGE_LIST_LOCK_DESTROY();
664 
665 	/* Callbacks may use the UMA zone. */
666 	NET_EPOCH_DRAIN_CALLBACKS();
667 
668 	uma_zdestroy(V_bridge_rtnode_zone);
669 }
670 VNET_SYSUNINIT(vnet_bridge_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY,
671     vnet_bridge_uninit, NULL);
672 
673 static int
674 bridge_modevent(module_t mod, int type, void *data)
675 {
676 
677 	switch (type) {
678 	case MOD_LOAD:
679 		bridge_dn_p = bridge_dummynet;
680 		bridge_same_p = bridge_same;
681 		bridge_get_softc_p = bridge_get_softc;
682 		bridge_member_ifaddrs_p = bridge_member_ifaddrs;
683 		bridge_detach_cookie = EVENTHANDLER_REGISTER(
684 		    ifnet_departure_event, bridge_ifdetach, NULL,
685 		    EVENTHANDLER_PRI_ANY);
686 		break;
687 	case MOD_UNLOAD:
688 		EVENTHANDLER_DEREGISTER(ifnet_departure_event,
689 		    bridge_detach_cookie);
690 		bridge_dn_p = NULL;
691 		bridge_same_p = NULL;
692 		bridge_get_softc_p = NULL;
693 		bridge_member_ifaddrs_p = NULL;
694 		break;
695 	default:
696 		return (EOPNOTSUPP);
697 	}
698 	return (0);
699 }
700 
701 static moduledata_t bridge_mod = {
702 	"if_bridge",
703 	bridge_modevent,
704 	0
705 };
706 
707 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
708 MODULE_VERSION(if_bridge, 1);
709 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
710 
711 /*
712  * handler for net.link.bridge.ipfw
713  */
714 static int
715 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
716 {
717 	int enable = V_pfil_ipfw;
718 	int error;
719 
720 	error = sysctl_handle_int(oidp, &enable, 0, req);
721 	enable &= 1;
722 
723 	if (enable != V_pfil_ipfw) {
724 		V_pfil_ipfw = enable;
725 
726 		/*
727 		 * Disable pfil so that ipfw doesnt run twice, if the user
728 		 * really wants both then they can re-enable pfil_bridge and/or
729 		 * pfil_member. Also allow non-ip packets as ipfw can filter by
730 		 * layer2 type.
731 		 */
732 		if (V_pfil_ipfw) {
733 			V_pfil_onlyip = 0;
734 			V_pfil_bridge = 0;
735 			V_pfil_member = 0;
736 		}
737 	}
738 
739 	return (error);
740 }
741 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw,
742     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_VNET | CTLFLAG_NEEDGIANT,
743     &VNET_NAME(pfil_ipfw), 0, &sysctl_pfil_ipfw, "I",
744     "Layer2 filter with IPFW");
745 
746 #ifdef VIMAGE
747 static void
748 bridge_reassign(struct ifnet *ifp, struct vnet *newvnet, char *arg)
749 {
750 	struct bridge_softc *sc = ifp->if_softc;
751 	struct bridge_iflist *bif;
752 
753 	BRIDGE_LOCK(sc);
754 
755 	while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
756 		bridge_delete_member(sc, bif, 0);
757 
758 	while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
759 		bridge_delete_span(sc, bif);
760 	}
761 
762 	BRIDGE_UNLOCK(sc);
763 
764 	ether_reassign(ifp, newvnet, arg);
765 }
766 #endif
767 
768 /*
769  * bridge_get_softc:
770  *
771  * Return the bridge softc for an ifnet.
772  */
773 static void *
774 bridge_get_softc(struct ifnet *ifp)
775 {
776 	struct bridge_iflist *bif;
777 
778 	NET_EPOCH_ASSERT();
779 
780 	bif = ifp->if_bridge;
781 	if (bif == NULL)
782 		return (NULL);
783 	return (bif->bif_sc);
784 }
785 
786 /*
787  * bridge_same:
788  *
789  * Return true if two interfaces are in the same bridge.  This is only used by
790  * bridgestp via bridge_same_p.
791  */
792 static bool
793 bridge_same(const void *bifap, const void *bifbp)
794 {
795 	const struct bridge_iflist *bifa = bifap, *bifb = bifbp;
796 
797 	NET_EPOCH_ASSERT();
798 
799 	if (bifa == NULL || bifb == NULL)
800 		return (false);
801 
802 	return (bifa->bif_sc == bifb->bif_sc);
803 }
804 
805 /*
806  * bridge_clone_create:
807  *
808  *	Create a new bridge instance.
809  */
810 static int
811 bridge_clone_create(struct if_clone *ifc, char *name, size_t len,
812     struct ifc_data *ifd, struct ifnet **ifpp)
813 {
814 	struct bridge_softc *sc;
815 	struct ifnet *ifp;
816 
817 	sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
818 	ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
819 
820 	BRIDGE_LOCK_INIT(sc);
821 	sc->sc_brtmax = BRIDGE_RTABLE_MAX;
822 	sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
823 
824 	/* Initialize our routing table. */
825 	bridge_rtable_init(sc);
826 
827 	callout_init_mtx(&sc->sc_brcallout, &sc->sc_rt_mtx, 0);
828 
829 	CK_LIST_INIT(&sc->sc_iflist);
830 	CK_LIST_INIT(&sc->sc_spanlist);
831 
832 	ifp->if_softc = sc;
833 	if_initname(ifp, bridge_name, ifd->unit);
834 	ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
835 	ifp->if_ioctl = bridge_ioctl;
836 #ifdef ALTQ
837 	ifp->if_start = bridge_altq_start;
838 	ifp->if_transmit = bridge_altq_transmit;
839 	IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
840 	ifp->if_snd.ifq_drv_maxlen = 0;
841 	IFQ_SET_READY(&ifp->if_snd);
842 #else
843 	ifp->if_transmit = bridge_transmit;
844 #endif
845 	ifp->if_qflush = bridge_qflush;
846 	ifp->if_init = bridge_init;
847 	ifp->if_type = IFT_BRIDGE;
848 
849 	ether_gen_addr(ifp, &sc->sc_defaddr);
850 
851 	bstp_attach(&sc->sc_stp, &bridge_ops);
852 	ether_ifattach(ifp, sc->sc_defaddr.octet);
853 	/* Now undo some of the damage... */
854 	ifp->if_baudrate = 0;
855 	ifp->if_type = IFT_BRIDGE;
856 #ifdef VIMAGE
857 	ifp->if_reassign = bridge_reassign;
858 #endif
859 	sc->sc_if_input = ifp->if_input;	/* ether_input */
860 	ifp->if_input = bridge_inject;
861 
862 	/*
863 	 * Allow BRIDGE_INPUT() to pass in packets originating from the bridge
864 	 * itself via bridge_inject().  This is required for netmap but
865 	 * otherwise has no effect.
866 	 */
867 	ifp->if_bridge_input = bridge_input;
868 
869 	BRIDGE_LIST_LOCK();
870 	LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list);
871 	BRIDGE_LIST_UNLOCK();
872 	*ifpp = ifp;
873 
874 	return (0);
875 }
876 
877 static void
878 bridge_clone_destroy_cb(struct epoch_context *ctx)
879 {
880 	struct bridge_softc *sc;
881 
882 	sc = __containerof(ctx, struct bridge_softc, sc_epoch_ctx);
883 
884 	BRIDGE_LOCK_DESTROY(sc);
885 	free(sc, M_DEVBUF);
886 }
887 
888 /*
889  * bridge_clone_destroy:
890  *
891  *	Destroy a bridge instance.
892  */
893 static int
894 bridge_clone_destroy(struct if_clone *ifc, struct ifnet *ifp, uint32_t flags)
895 {
896 	struct bridge_softc *sc = ifp->if_softc;
897 	struct bridge_iflist *bif;
898 	struct epoch_tracker et;
899 
900 	BRIDGE_LOCK(sc);
901 
902 	bridge_stop(ifp, 1);
903 	ifp->if_flags &= ~IFF_UP;
904 
905 	while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
906 		bridge_delete_member(sc, bif, 0);
907 
908 	while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
909 		bridge_delete_span(sc, bif);
910 	}
911 
912 	/* Tear down the routing table. */
913 	bridge_rtable_fini(sc);
914 
915 	BRIDGE_UNLOCK(sc);
916 
917 	NET_EPOCH_ENTER(et);
918 
919 	callout_drain(&sc->sc_brcallout);
920 
921 	BRIDGE_LIST_LOCK();
922 	LIST_REMOVE(sc, sc_list);
923 	BRIDGE_LIST_UNLOCK();
924 
925 	bstp_detach(&sc->sc_stp);
926 #ifdef ALTQ
927 	IFQ_PURGE(&ifp->if_snd);
928 #endif
929 	NET_EPOCH_EXIT(et);
930 
931 	ether_ifdetach(ifp);
932 	if_free(ifp);
933 
934 	NET_EPOCH_CALL(bridge_clone_destroy_cb, &sc->sc_epoch_ctx);
935 
936 	return (0);
937 }
938 
939 /*
940  * bridge_ioctl:
941  *
942  *	Handle a control request from the operator.
943  */
944 static int
945 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
946 {
947 	struct bridge_softc *sc = ifp->if_softc;
948 	struct ifreq *ifr = (struct ifreq *)data;
949 	struct bridge_iflist *bif;
950 	struct thread *td = curthread;
951 	union {
952 		struct ifbreq ifbreq;
953 		struct ifbifconf ifbifconf;
954 		struct ifbareq ifbareq;
955 		struct ifbaconf ifbaconf;
956 		struct ifbrparam ifbrparam;
957 		struct ifbropreq ifbropreq;
958 	} args;
959 	struct ifdrv *ifd = (struct ifdrv *) data;
960 	const struct bridge_control *bc;
961 	int error = 0, oldmtu;
962 
963 	BRIDGE_LOCK(sc);
964 
965 	switch (cmd) {
966 	case SIOCADDMULTI:
967 	case SIOCDELMULTI:
968 		break;
969 
970 	case SIOCGDRVSPEC:
971 	case SIOCSDRVSPEC:
972 		if (ifd->ifd_cmd >= bridge_control_table_size) {
973 			error = EINVAL;
974 			break;
975 		}
976 		bc = &bridge_control_table[ifd->ifd_cmd];
977 
978 		if (cmd == SIOCGDRVSPEC &&
979 		    (bc->bc_flags & BC_F_COPYOUT) == 0) {
980 			error = EINVAL;
981 			break;
982 		}
983 		else if (cmd == SIOCSDRVSPEC &&
984 		    (bc->bc_flags & BC_F_COPYOUT) != 0) {
985 			error = EINVAL;
986 			break;
987 		}
988 
989 		if (bc->bc_flags & BC_F_SUSER) {
990 			error = priv_check(td, PRIV_NET_BRIDGE);
991 			if (error)
992 				break;
993 		}
994 
995 		if (ifd->ifd_len != bc->bc_argsize ||
996 		    ifd->ifd_len > sizeof(args)) {
997 			error = EINVAL;
998 			break;
999 		}
1000 
1001 		bzero(&args, sizeof(args));
1002 		if (bc->bc_flags & BC_F_COPYIN) {
1003 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
1004 			if (error)
1005 				break;
1006 		}
1007 
1008 		oldmtu = ifp->if_mtu;
1009 		error = (*bc->bc_func)(sc, &args);
1010 		if (error)
1011 			break;
1012 
1013 		/*
1014 		 * Bridge MTU may change during addition of the first port.
1015 		 * If it did, do network layer specific procedure.
1016 		 */
1017 		if (ifp->if_mtu != oldmtu)
1018 			if_notifymtu(ifp);
1019 
1020 		if (bc->bc_flags & BC_F_COPYOUT)
1021 			error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
1022 
1023 		break;
1024 
1025 	case SIOCSIFFLAGS:
1026 		if (!(ifp->if_flags & IFF_UP) &&
1027 		    (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
1028 			/*
1029 			 * If interface is marked down and it is running,
1030 			 * then stop and disable it.
1031 			 */
1032 			bridge_stop(ifp, 1);
1033 		} else if ((ifp->if_flags & IFF_UP) &&
1034 		    !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
1035 			/*
1036 			 * If interface is marked up and it is stopped, then
1037 			 * start it.
1038 			 */
1039 			BRIDGE_UNLOCK(sc);
1040 			(*ifp->if_init)(sc);
1041 			BRIDGE_LOCK(sc);
1042 		}
1043 		break;
1044 
1045 	case SIOCSIFMTU:
1046 		oldmtu = sc->sc_ifp->if_mtu;
1047 
1048 		if (ifr->ifr_mtu < IF_MINMTU) {
1049 			error = EINVAL;
1050 			break;
1051 		}
1052 		if (CK_LIST_EMPTY(&sc->sc_iflist)) {
1053 			sc->sc_ifp->if_mtu = ifr->ifr_mtu;
1054 			break;
1055 		}
1056 		CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1057 			error = (*bif->bif_ifp->if_ioctl)(bif->bif_ifp,
1058 			    SIOCSIFMTU, (caddr_t)ifr);
1059 			if (error != 0) {
1060 				log(LOG_NOTICE, "%s: invalid MTU: %u for"
1061 				    " member %s\n", sc->sc_ifp->if_xname,
1062 				    ifr->ifr_mtu,
1063 				    bif->bif_ifp->if_xname);
1064 				error = EINVAL;
1065 				break;
1066 			}
1067 		}
1068 		if (error) {
1069 			/* Restore the previous MTU on all member interfaces. */
1070 			ifr->ifr_mtu = oldmtu;
1071 			CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1072 				(*bif->bif_ifp->if_ioctl)(bif->bif_ifp,
1073 				    SIOCSIFMTU, (caddr_t)ifr);
1074 			}
1075 		} else {
1076 			sc->sc_ifp->if_mtu = ifr->ifr_mtu;
1077 		}
1078 		break;
1079 	default:
1080 		/*
1081 		 * drop the lock as ether_ioctl() will call bridge_start() and
1082 		 * cause the lock to be recursed.
1083 		 */
1084 		BRIDGE_UNLOCK(sc);
1085 		error = ether_ioctl(ifp, cmd, data);
1086 		BRIDGE_LOCK(sc);
1087 		break;
1088 	}
1089 
1090 	BRIDGE_UNLOCK(sc);
1091 
1092 	return (error);
1093 }
1094 
1095 /*
1096  * bridge_mutecaps:
1097  *
1098  *	Clear or restore unwanted capabilities on the member interface
1099  */
1100 static void
1101 bridge_mutecaps(struct bridge_softc *sc)
1102 {
1103 	struct bridge_iflist *bif;
1104 	int enabled, mask;
1105 
1106 	BRIDGE_LOCK_ASSERT(sc);
1107 
1108 	/* Initial bitmask of capabilities to test */
1109 	mask = BRIDGE_IFCAPS_MASK;
1110 
1111 	CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1112 		/* Every member must support it or its disabled */
1113 		mask &= bif->bif_savedcaps;
1114 	}
1115 
1116 	CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1117 		enabled = bif->bif_ifp->if_capenable;
1118 		enabled &= ~BRIDGE_IFCAPS_STRIP;
1119 		/* strip off mask bits and enable them again if allowed */
1120 		enabled &= ~BRIDGE_IFCAPS_MASK;
1121 		enabled |= mask;
1122 		bridge_set_ifcap(sc, bif, enabled);
1123 	}
1124 }
1125 
1126 static void
1127 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
1128 {
1129 	struct ifnet *ifp = bif->bif_ifp;
1130 	struct ifreq ifr;
1131 	int error, mask, stuck;
1132 
1133 	bzero(&ifr, sizeof(ifr));
1134 	ifr.ifr_reqcap = set;
1135 
1136 	if (ifp->if_capenable != set) {
1137 		error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
1138 		if (error)
1139 			if_printf(sc->sc_ifp,
1140 			    "error setting capabilities on %s: %d\n",
1141 			    ifp->if_xname, error);
1142 		mask = BRIDGE_IFCAPS_MASK | BRIDGE_IFCAPS_STRIP;
1143 		stuck = ifp->if_capenable & mask & ~set;
1144 		if (stuck != 0)
1145 			if_printf(sc->sc_ifp,
1146 			    "can't disable some capabilities on %s: 0x%x\n",
1147 			    ifp->if_xname, stuck);
1148 	}
1149 }
1150 
1151 /*
1152  * bridge_lookup_member:
1153  *
1154  *	Lookup a bridge member interface.
1155  */
1156 static struct bridge_iflist *
1157 bridge_lookup_member(struct bridge_softc *sc, const char *name)
1158 {
1159 	struct bridge_iflist *bif;
1160 	struct ifnet *ifp;
1161 
1162 	BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1163 
1164 	CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1165 		ifp = bif->bif_ifp;
1166 		if (strcmp(ifp->if_xname, name) == 0)
1167 			return (bif);
1168 	}
1169 
1170 	return (NULL);
1171 }
1172 
1173 /*
1174  * bridge_lookup_member_if:
1175  *
1176  *	Lookup a bridge member interface by ifnet*.
1177  */
1178 static struct bridge_iflist *
1179 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
1180 {
1181 	BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1182 	return (member_ifp->if_bridge);
1183 }
1184 
1185 static void
1186 bridge_delete_member_cb(struct epoch_context *ctx)
1187 {
1188 	struct bridge_iflist *bif;
1189 
1190 	bif = __containerof(ctx, struct bridge_iflist, bif_epoch_ctx);
1191 
1192 	free(bif, M_DEVBUF);
1193 }
1194 
1195 /*
1196  * bridge_delete_member:
1197  *
1198  *	Delete the specified member interface.
1199  */
1200 static void
1201 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
1202     int gone)
1203 {
1204 	struct ifnet *ifs = bif->bif_ifp;
1205 	struct ifnet *fif = NULL;
1206 	struct bridge_iflist *bifl;
1207 
1208 	BRIDGE_LOCK_ASSERT(sc);
1209 
1210 	if (bif->bif_flags & IFBIF_STP)
1211 		bstp_disable(&bif->bif_stp);
1212 
1213 	ifs->if_bridge = NULL;
1214 	CK_LIST_REMOVE(bif, bif_next);
1215 
1216 	/*
1217 	 * If removing the interface that gave the bridge its mac address, set
1218 	 * the mac address of the bridge to the address of the next member, or
1219 	 * to its default address if no members are left.
1220 	 */
1221 	if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) {
1222 		if (CK_LIST_EMPTY(&sc->sc_iflist)) {
1223 			bcopy(&sc->sc_defaddr,
1224 			    IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1225 			sc->sc_ifaddr = NULL;
1226 		} else {
1227 			bifl = CK_LIST_FIRST(&sc->sc_iflist);
1228 			fif = bifl->bif_ifp;
1229 			bcopy(IF_LLADDR(fif),
1230 			    IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1231 			sc->sc_ifaddr = fif;
1232 		}
1233 		EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1234 	}
1235 
1236 	bridge_linkcheck(sc);
1237 	bridge_mutecaps(sc);	/* recalcuate now this interface is removed */
1238 	BRIDGE_RT_LOCK(sc);
1239 	bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1240 	BRIDGE_RT_UNLOCK(sc);
1241 	KASSERT(bif->bif_addrcnt == 0,
1242 	    ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
1243 
1244 	ifs->if_bridge_output = NULL;
1245 	ifs->if_bridge_input = NULL;
1246 	ifs->if_bridge_linkstate = NULL;
1247 	if (!gone) {
1248 		switch (ifs->if_type) {
1249 		case IFT_ETHER:
1250 		case IFT_L2VLAN:
1251 			/*
1252 			 * Take the interface out of promiscuous mode, but only
1253 			 * if it was promiscuous in the first place. It might
1254 			 * not be if we're in the bridge_ioctl_add() error path.
1255 			 */
1256 			if (ifs->if_flags & IFF_PROMISC)
1257 				(void) ifpromisc(ifs, 0);
1258 			break;
1259 
1260 		case IFT_GIF:
1261 			break;
1262 
1263 		default:
1264 #ifdef DIAGNOSTIC
1265 			panic("bridge_delete_member: impossible");
1266 #endif
1267 			break;
1268 		}
1269 		/* reneable any interface capabilities */
1270 		bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1271 	}
1272 	bstp_destroy(&bif->bif_stp);	/* prepare to free */
1273 
1274 	NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1275 }
1276 
1277 /*
1278  * bridge_delete_span:
1279  *
1280  *	Delete the specified span interface.
1281  */
1282 static void
1283 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1284 {
1285 	BRIDGE_LOCK_ASSERT(sc);
1286 
1287 	KASSERT(bif->bif_ifp->if_bridge == NULL,
1288 	    ("%s: not a span interface", __func__));
1289 
1290 	CK_LIST_REMOVE(bif, bif_next);
1291 
1292 	NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1293 }
1294 
1295 static int
1296 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1297 {
1298 	struct ifbreq *req = arg;
1299 	struct bridge_iflist *bif = NULL;
1300 	struct ifnet *ifs;
1301 	int error = 0;
1302 
1303 	ifs = ifunit(req->ifbr_ifsname);
1304 	if (ifs == NULL)
1305 		return (ENOENT);
1306 	if (ifs->if_ioctl == NULL)	/* must be supported */
1307 		return (EINVAL);
1308 
1309 	/* If it's in the span list, it can't be a member. */
1310 	CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1311 		if (ifs == bif->bif_ifp)
1312 			return (EBUSY);
1313 
1314 	if (ifs->if_bridge) {
1315 		struct bridge_iflist *sbif = ifs->if_bridge;
1316 		if (sbif->bif_sc == sc)
1317 			return (EEXIST);
1318 
1319 		return (EBUSY);
1320 	}
1321 
1322 	switch (ifs->if_type) {
1323 	case IFT_ETHER:
1324 	case IFT_L2VLAN:
1325 	case IFT_GIF:
1326 		/* permitted interface types */
1327 		break;
1328 	default:
1329 		return (EINVAL);
1330 	}
1331 
1332 	/*
1333 	 * If member_ifaddrs is disabled, do not allow an interface with
1334 	 * assigned IP addresses to be added to a bridge.
1335 	 */
1336 	if (!V_member_ifaddrs) {
1337 		struct ifaddr *ifa;
1338 
1339 		CK_STAILQ_FOREACH(ifa, &ifs->if_addrhead, ifa_link) {
1340 #ifdef INET
1341 			if (ifa->ifa_addr->sa_family == AF_INET)
1342 				return (EINVAL);
1343 #endif
1344 #ifdef INET6
1345 			if (ifa->ifa_addr->sa_family == AF_INET6)
1346 				return (EINVAL);
1347 #endif
1348 		}
1349 	}
1350 
1351 #ifdef INET6
1352 	/*
1353 	 * Two valid inet6 addresses with link-local scope must not be
1354 	 * on the parent interface and the member interfaces at the
1355 	 * same time.  This restriction is needed to prevent violation
1356 	 * of link-local scope zone.  Attempts to add a member
1357 	 * interface which has inet6 addresses when the parent has
1358 	 * inet6 triggers removal of all inet6 addresses on the member
1359 	 * interface.
1360 	 */
1361 
1362 	/* Check if the parent interface has a link-local scope addr. */
1363 	if (V_allow_llz_overlap == 0 &&
1364 	    in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1365 		/*
1366 		 * If any, remove all inet6 addresses from the member
1367 		 * interfaces.
1368 		 */
1369 		CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1370  			if (in6ifa_llaonifp(bif->bif_ifp)) {
1371 				in6_ifdetach(bif->bif_ifp);
1372 				if_printf(sc->sc_ifp,
1373 				    "IPv6 addresses on %s have been removed "
1374 				    "before adding it as a member to prevent "
1375 				    "IPv6 address scope violation.\n",
1376 				    bif->bif_ifp->if_xname);
1377 			}
1378 		}
1379 		if (in6ifa_llaonifp(ifs)) {
1380 			in6_ifdetach(ifs);
1381 			if_printf(sc->sc_ifp,
1382 			    "IPv6 addresses on %s have been removed "
1383 			    "before adding it as a member to prevent "
1384 			    "IPv6 address scope violation.\n",
1385 			    ifs->if_xname);
1386 		}
1387 	}
1388 #endif
1389 	/* Allow the first Ethernet member to define the MTU */
1390 	if (CK_LIST_EMPTY(&sc->sc_iflist))
1391 		sc->sc_ifp->if_mtu = ifs->if_mtu;
1392 	else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1393 		struct ifreq ifr;
1394 
1395 		snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "%s",
1396 		    ifs->if_xname);
1397 		ifr.ifr_mtu = sc->sc_ifp->if_mtu;
1398 
1399 		error = (*ifs->if_ioctl)(ifs,
1400 		    SIOCSIFMTU, (caddr_t)&ifr);
1401 		if (error != 0) {
1402 			log(LOG_NOTICE, "%s: invalid MTU: %u for"
1403 			    " new member %s\n", sc->sc_ifp->if_xname,
1404 			    ifr.ifr_mtu,
1405 			    ifs->if_xname);
1406 			return (EINVAL);
1407 		}
1408 	}
1409 
1410 	bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1411 	if (bif == NULL)
1412 		return (ENOMEM);
1413 
1414 	bif->bif_sc = sc;
1415 	bif->bif_ifp = ifs;
1416 	bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1417 	bif->bif_savedcaps = ifs->if_capenable;
1418 
1419 	/*
1420 	 * Assign the interface's MAC address to the bridge if it's the first
1421 	 * member and the MAC address of the bridge has not been changed from
1422 	 * the default randomly generated one.
1423 	 */
1424 	if (V_bridge_inherit_mac && CK_LIST_EMPTY(&sc->sc_iflist) &&
1425 	    !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr.octet, ETHER_ADDR_LEN)) {
1426 		bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1427 		sc->sc_ifaddr = ifs;
1428 		EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1429 	}
1430 
1431 	ifs->if_bridge = bif;
1432 	ifs->if_bridge_output = bridge_output;
1433 	ifs->if_bridge_input = bridge_input;
1434 	ifs->if_bridge_linkstate = bridge_linkstate;
1435 	bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1436 	/*
1437 	 * XXX: XLOCK HERE!?!
1438 	 *
1439 	 * NOTE: insert_***HEAD*** should be safe for the traversals.
1440 	 */
1441 	CK_LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1442 
1443 	/* Set interface capabilities to the intersection set of all members */
1444 	bridge_mutecaps(sc);
1445 	bridge_linkcheck(sc);
1446 
1447 	/* Place the interface into promiscuous mode */
1448 	switch (ifs->if_type) {
1449 		case IFT_ETHER:
1450 		case IFT_L2VLAN:
1451 			error = ifpromisc(ifs, 1);
1452 			break;
1453 	}
1454 
1455 	if (error)
1456 		bridge_delete_member(sc, bif, 0);
1457 	return (error);
1458 }
1459 
1460 static int
1461 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1462 {
1463 	struct ifbreq *req = arg;
1464 	struct bridge_iflist *bif;
1465 
1466 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1467 	if (bif == NULL)
1468 		return (ENOENT);
1469 
1470 	bridge_delete_member(sc, bif, 0);
1471 
1472 	return (0);
1473 }
1474 
1475 static int
1476 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1477 {
1478 	struct ifbreq *req = arg;
1479 	struct bridge_iflist *bif;
1480 	struct bstp_port *bp;
1481 
1482 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1483 	if (bif == NULL)
1484 		return (ENOENT);
1485 
1486 	bp = &bif->bif_stp;
1487 	req->ifbr_ifsflags = bif->bif_flags;
1488 	req->ifbr_state = bp->bp_state;
1489 	req->ifbr_priority = bp->bp_priority;
1490 	req->ifbr_path_cost = bp->bp_path_cost;
1491 	req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1492 	req->ifbr_proto = bp->bp_protover;
1493 	req->ifbr_role = bp->bp_role;
1494 	req->ifbr_stpflags = bp->bp_flags;
1495 	req->ifbr_addrcnt = bif->bif_addrcnt;
1496 	req->ifbr_addrmax = bif->bif_addrmax;
1497 	req->ifbr_addrexceeded = bif->bif_addrexceeded;
1498 
1499 	/* Copy STP state options as flags */
1500 	if (bp->bp_operedge)
1501 		req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1502 	if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1503 		req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1504 	if (bp->bp_ptp_link)
1505 		req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1506 	if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1507 		req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1508 	if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1509 		req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1510 	if (bp->bp_flags & BSTP_PORT_ADMCOST)
1511 		req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1512 	return (0);
1513 }
1514 
1515 static int
1516 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1517 {
1518 	struct epoch_tracker et;
1519 	struct ifbreq *req = arg;
1520 	struct bridge_iflist *bif;
1521 	struct bstp_port *bp;
1522 	int error;
1523 
1524 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1525 	if (bif == NULL)
1526 		return (ENOENT);
1527 	bp = &bif->bif_stp;
1528 
1529 	if (req->ifbr_ifsflags & IFBIF_SPAN)
1530 		/* SPAN is readonly */
1531 		return (EINVAL);
1532 
1533 	NET_EPOCH_ENTER(et);
1534 
1535 	if (req->ifbr_ifsflags & IFBIF_STP) {
1536 		if ((bif->bif_flags & IFBIF_STP) == 0) {
1537 			error = bstp_enable(&bif->bif_stp);
1538 			if (error) {
1539 				NET_EPOCH_EXIT(et);
1540 				return (error);
1541 			}
1542 		}
1543 	} else {
1544 		if ((bif->bif_flags & IFBIF_STP) != 0)
1545 			bstp_disable(&bif->bif_stp);
1546 	}
1547 
1548 	/* Pass on STP flags */
1549 	bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1550 	bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1551 	bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1552 	bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1553 
1554 	/* Save the bits relating to the bridge */
1555 	bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1556 
1557 	NET_EPOCH_EXIT(et);
1558 
1559 	return (0);
1560 }
1561 
1562 static int
1563 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1564 {
1565 	struct ifbrparam *param = arg;
1566 
1567 	sc->sc_brtmax = param->ifbrp_csize;
1568 	bridge_rttrim(sc);
1569 
1570 	return (0);
1571 }
1572 
1573 static int
1574 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1575 {
1576 	struct ifbrparam *param = arg;
1577 
1578 	param->ifbrp_csize = sc->sc_brtmax;
1579 
1580 	return (0);
1581 }
1582 
1583 static int
1584 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1585 {
1586 	struct ifbifconf *bifc = arg;
1587 	struct bridge_iflist *bif;
1588 	struct ifbreq breq;
1589 	char *buf, *outbuf;
1590 	int count, buflen, len, error = 0;
1591 
1592 	count = 0;
1593 	CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1594 		count++;
1595 	CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1596 		count++;
1597 
1598 	buflen = sizeof(breq) * count;
1599 	if (bifc->ifbic_len == 0) {
1600 		bifc->ifbic_len = buflen;
1601 		return (0);
1602 	}
1603 	outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1604 	if (outbuf == NULL)
1605 		return (ENOMEM);
1606 
1607 	count = 0;
1608 	buf = outbuf;
1609 	len = min(bifc->ifbic_len, buflen);
1610 	bzero(&breq, sizeof(breq));
1611 	CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1612 		if (len < sizeof(breq))
1613 			break;
1614 
1615 		strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1616 		    sizeof(breq.ifbr_ifsname));
1617 		/* Fill in the ifbreq structure */
1618 		error = bridge_ioctl_gifflags(sc, &breq);
1619 		if (error)
1620 			break;
1621 		memcpy(buf, &breq, sizeof(breq));
1622 		count++;
1623 		buf += sizeof(breq);
1624 		len -= sizeof(breq);
1625 	}
1626 	CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1627 		if (len < sizeof(breq))
1628 			break;
1629 
1630 		strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1631 		    sizeof(breq.ifbr_ifsname));
1632 		breq.ifbr_ifsflags = bif->bif_flags;
1633 		breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1634 		memcpy(buf, &breq, sizeof(breq));
1635 		count++;
1636 		buf += sizeof(breq);
1637 		len -= sizeof(breq);
1638 	}
1639 
1640 	bifc->ifbic_len = sizeof(breq) * count;
1641 	error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1642 	free(outbuf, M_TEMP);
1643 	return (error);
1644 }
1645 
1646 static int
1647 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1648 {
1649 	struct ifbaconf *bac = arg;
1650 	struct bridge_rtnode *brt;
1651 	struct ifbareq bareq;
1652 	char *buf, *outbuf;
1653 	int count, buflen, len, error = 0;
1654 
1655 	if (bac->ifbac_len == 0)
1656 		return (0);
1657 
1658 	count = 0;
1659 	CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1660 		count++;
1661 	buflen = sizeof(bareq) * count;
1662 
1663 	outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1664 	if (outbuf == NULL)
1665 		return (ENOMEM);
1666 
1667 	count = 0;
1668 	buf = outbuf;
1669 	len = min(bac->ifbac_len, buflen);
1670 	bzero(&bareq, sizeof(bareq));
1671 	CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1672 		if (len < sizeof(bareq))
1673 			goto out;
1674 		strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1675 		    sizeof(bareq.ifba_ifsname));
1676 		memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1677 		bareq.ifba_vlan = brt->brt_vlan;
1678 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1679 				time_uptime < brt->brt_expire)
1680 			bareq.ifba_expire = brt->brt_expire - time_uptime;
1681 		else
1682 			bareq.ifba_expire = 0;
1683 		bareq.ifba_flags = brt->brt_flags;
1684 
1685 		memcpy(buf, &bareq, sizeof(bareq));
1686 		count++;
1687 		buf += sizeof(bareq);
1688 		len -= sizeof(bareq);
1689 	}
1690 out:
1691 	bac->ifbac_len = sizeof(bareq) * count;
1692 	error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1693 	free(outbuf, M_TEMP);
1694 	return (error);
1695 }
1696 
1697 static int
1698 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1699 {
1700 	struct ifbareq *req = arg;
1701 	struct bridge_iflist *bif;
1702 	struct epoch_tracker et;
1703 	int error;
1704 
1705 	NET_EPOCH_ENTER(et);
1706 	bif = bridge_lookup_member(sc, req->ifba_ifsname);
1707 	if (bif == NULL) {
1708 		NET_EPOCH_EXIT(et);
1709 		return (ENOENT);
1710 	}
1711 
1712 	/* bridge_rtupdate() may acquire the lock. */
1713 	error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1714 	    req->ifba_flags);
1715 	NET_EPOCH_EXIT(et);
1716 
1717 	return (error);
1718 }
1719 
1720 static int
1721 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1722 {
1723 	struct ifbrparam *param = arg;
1724 
1725 	sc->sc_brttimeout = param->ifbrp_ctime;
1726 	return (0);
1727 }
1728 
1729 static int
1730 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1731 {
1732 	struct ifbrparam *param = arg;
1733 
1734 	param->ifbrp_ctime = sc->sc_brttimeout;
1735 	return (0);
1736 }
1737 
1738 static int
1739 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1740 {
1741 	struct ifbareq *req = arg;
1742 	int vlan = req->ifba_vlan;
1743 
1744 	/* Userspace uses '0' to mean 'any vlan' */
1745 	if (vlan == 0)
1746 		vlan = DOT1Q_VID_RSVD_IMPL;
1747 
1748 	return (bridge_rtdaddr(sc, req->ifba_dst, vlan));
1749 }
1750 
1751 static int
1752 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1753 {
1754 	struct ifbreq *req = arg;
1755 
1756 	BRIDGE_RT_LOCK(sc);
1757 	bridge_rtflush(sc, req->ifbr_ifsflags);
1758 	BRIDGE_RT_UNLOCK(sc);
1759 
1760 	return (0);
1761 }
1762 
1763 static int
1764 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1765 {
1766 	struct ifbrparam *param = arg;
1767 	struct bstp_state *bs = &sc->sc_stp;
1768 
1769 	param->ifbrp_prio = bs->bs_bridge_priority;
1770 	return (0);
1771 }
1772 
1773 static int
1774 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1775 {
1776 	struct ifbrparam *param = arg;
1777 
1778 	return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1779 }
1780 
1781 static int
1782 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1783 {
1784 	struct ifbrparam *param = arg;
1785 	struct bstp_state *bs = &sc->sc_stp;
1786 
1787 	param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1788 	return (0);
1789 }
1790 
1791 static int
1792 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1793 {
1794 	struct ifbrparam *param = arg;
1795 
1796 	return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1797 }
1798 
1799 static int
1800 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1801 {
1802 	struct ifbrparam *param = arg;
1803 	struct bstp_state *bs = &sc->sc_stp;
1804 
1805 	param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1806 	return (0);
1807 }
1808 
1809 static int
1810 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1811 {
1812 	struct ifbrparam *param = arg;
1813 
1814 	return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1815 }
1816 
1817 static int
1818 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1819 {
1820 	struct ifbrparam *param = arg;
1821 	struct bstp_state *bs = &sc->sc_stp;
1822 
1823 	param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1824 	return (0);
1825 }
1826 
1827 static int
1828 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1829 {
1830 	struct ifbrparam *param = arg;
1831 
1832 	return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1833 }
1834 
1835 static int
1836 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1837 {
1838 	struct ifbreq *req = arg;
1839 	struct bridge_iflist *bif;
1840 
1841 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1842 	if (bif == NULL)
1843 		return (ENOENT);
1844 
1845 	return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1846 }
1847 
1848 static int
1849 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1850 {
1851 	struct ifbreq *req = arg;
1852 	struct bridge_iflist *bif;
1853 
1854 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1855 	if (bif == NULL)
1856 		return (ENOENT);
1857 
1858 	return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1859 }
1860 
1861 static int
1862 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1863 {
1864 	struct ifbreq *req = arg;
1865 	struct bridge_iflist *bif;
1866 
1867 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1868 	if (bif == NULL)
1869 		return (ENOENT);
1870 
1871 	bif->bif_addrmax = req->ifbr_addrmax;
1872 	return (0);
1873 }
1874 
1875 static int
1876 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1877 {
1878 	struct ifbreq *req = arg;
1879 	struct bridge_iflist *bif = NULL;
1880 	struct ifnet *ifs;
1881 
1882 	ifs = ifunit(req->ifbr_ifsname);
1883 	if (ifs == NULL)
1884 		return (ENOENT);
1885 
1886 	CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1887 		if (ifs == bif->bif_ifp)
1888 			return (EBUSY);
1889 
1890 	if (ifs->if_bridge != NULL)
1891 		return (EBUSY);
1892 
1893 	switch (ifs->if_type) {
1894 		case IFT_ETHER:
1895 		case IFT_GIF:
1896 		case IFT_L2VLAN:
1897 			break;
1898 		default:
1899 			return (EINVAL);
1900 	}
1901 
1902 	bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1903 	if (bif == NULL)
1904 		return (ENOMEM);
1905 
1906 	bif->bif_ifp = ifs;
1907 	bif->bif_flags = IFBIF_SPAN;
1908 
1909 	CK_LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1910 
1911 	return (0);
1912 }
1913 
1914 static int
1915 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1916 {
1917 	struct ifbreq *req = arg;
1918 	struct bridge_iflist *bif;
1919 	struct ifnet *ifs;
1920 
1921 	ifs = ifunit(req->ifbr_ifsname);
1922 	if (ifs == NULL)
1923 		return (ENOENT);
1924 
1925 	CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1926 		if (ifs == bif->bif_ifp)
1927 			break;
1928 
1929 	if (bif == NULL)
1930 		return (ENOENT);
1931 
1932 	bridge_delete_span(sc, bif);
1933 
1934 	return (0);
1935 }
1936 
1937 static int
1938 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1939 {
1940 	struct ifbropreq *req = arg;
1941 	struct bstp_state *bs = &sc->sc_stp;
1942 	struct bstp_port *root_port;
1943 
1944 	req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1945 	req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1946 	req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1947 
1948 	root_port = bs->bs_root_port;
1949 	if (root_port == NULL)
1950 		req->ifbop_root_port = 0;
1951 	else
1952 		req->ifbop_root_port = root_port->bp_ifp->if_index;
1953 
1954 	req->ifbop_holdcount = bs->bs_txholdcount;
1955 	req->ifbop_priority = bs->bs_bridge_priority;
1956 	req->ifbop_protocol = bs->bs_protover;
1957 	req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1958 	req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1959 	req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1960 	req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1961 	req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1962 	req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1963 
1964 	return (0);
1965 }
1966 
1967 static int
1968 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1969 {
1970 	struct ifbrparam *param = arg;
1971 
1972 	param->ifbrp_cexceeded = sc->sc_brtexceeded;
1973 	return (0);
1974 }
1975 
1976 static int
1977 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1978 {
1979 	struct ifbpstpconf *bifstp = arg;
1980 	struct bridge_iflist *bif;
1981 	struct bstp_port *bp;
1982 	struct ifbpstpreq bpreq;
1983 	char *buf, *outbuf;
1984 	int count, buflen, len, error = 0;
1985 
1986 	count = 0;
1987 	CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1988 		if ((bif->bif_flags & IFBIF_STP) != 0)
1989 			count++;
1990 	}
1991 
1992 	buflen = sizeof(bpreq) * count;
1993 	if (bifstp->ifbpstp_len == 0) {
1994 		bifstp->ifbpstp_len = buflen;
1995 		return (0);
1996 	}
1997 
1998 	outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1999 	if (outbuf == NULL)
2000 		return (ENOMEM);
2001 
2002 	count = 0;
2003 	buf = outbuf;
2004 	len = min(bifstp->ifbpstp_len, buflen);
2005 	bzero(&bpreq, sizeof(bpreq));
2006 	CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2007 		if (len < sizeof(bpreq))
2008 			break;
2009 
2010 		if ((bif->bif_flags & IFBIF_STP) == 0)
2011 			continue;
2012 
2013 		bp = &bif->bif_stp;
2014 		bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
2015 		bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
2016 		bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
2017 		bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
2018 		bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
2019 		bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
2020 
2021 		memcpy(buf, &bpreq, sizeof(bpreq));
2022 		count++;
2023 		buf += sizeof(bpreq);
2024 		len -= sizeof(bpreq);
2025 	}
2026 
2027 	bifstp->ifbpstp_len = sizeof(bpreq) * count;
2028 	error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
2029 	free(outbuf, M_TEMP);
2030 	return (error);
2031 }
2032 
2033 static int
2034 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
2035 {
2036 	struct ifbrparam *param = arg;
2037 
2038 	return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
2039 }
2040 
2041 static int
2042 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
2043 {
2044 	struct ifbrparam *param = arg;
2045 
2046 	return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
2047 }
2048 
2049 /*
2050  * bridge_ifdetach:
2051  *
2052  *	Detach an interface from a bridge.  Called when a member
2053  *	interface is detaching.
2054  */
2055 static void
2056 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
2057 {
2058 	struct bridge_iflist *bif = ifp->if_bridge;
2059 	struct bridge_softc *sc = NULL;
2060 
2061 	if (bif)
2062 		sc = bif->bif_sc;
2063 
2064 	if (ifp->if_flags & IFF_RENAMING)
2065 		return;
2066 	if (V_bridge_cloner == NULL) {
2067 		/*
2068 		 * This detach handler can be called after
2069 		 * vnet_bridge_uninit().  Just return in that case.
2070 		 */
2071 		return;
2072 	}
2073 	/* Check if the interface is a bridge member */
2074 	if (sc != NULL) {
2075 		BRIDGE_LOCK(sc);
2076 		bridge_delete_member(sc, bif, 1);
2077 		BRIDGE_UNLOCK(sc);
2078 		return;
2079 	}
2080 
2081 	/* Check if the interface is a span port */
2082 	BRIDGE_LIST_LOCK();
2083 	LIST_FOREACH(sc, &V_bridge_list, sc_list) {
2084 		BRIDGE_LOCK(sc);
2085 		CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
2086 			if (ifp == bif->bif_ifp) {
2087 				bridge_delete_span(sc, bif);
2088 				break;
2089 			}
2090 
2091 		BRIDGE_UNLOCK(sc);
2092 	}
2093 	BRIDGE_LIST_UNLOCK();
2094 }
2095 
2096 /*
2097  * bridge_init:
2098  *
2099  *	Initialize a bridge interface.
2100  */
2101 static void
2102 bridge_init(void *xsc)
2103 {
2104 	struct bridge_softc *sc = (struct bridge_softc *)xsc;
2105 	struct ifnet *ifp = sc->sc_ifp;
2106 
2107 	if (ifp->if_drv_flags & IFF_DRV_RUNNING)
2108 		return;
2109 
2110 	BRIDGE_LOCK(sc);
2111 	callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
2112 	    bridge_timer, sc);
2113 
2114 	ifp->if_drv_flags |= IFF_DRV_RUNNING;
2115 	bstp_init(&sc->sc_stp);		/* Initialize Spanning Tree */
2116 
2117 	BRIDGE_UNLOCK(sc);
2118 }
2119 
2120 /*
2121  * bridge_stop:
2122  *
2123  *	Stop the bridge interface.
2124  */
2125 static void
2126 bridge_stop(struct ifnet *ifp, int disable)
2127 {
2128 	struct bridge_softc *sc = ifp->if_softc;
2129 
2130 	BRIDGE_LOCK_ASSERT(sc);
2131 
2132 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2133 		return;
2134 
2135 	BRIDGE_RT_LOCK(sc);
2136 	callout_stop(&sc->sc_brcallout);
2137 
2138 	bstp_stop(&sc->sc_stp);
2139 
2140 	bridge_rtflush(sc, IFBF_FLUSHDYN);
2141 	BRIDGE_RT_UNLOCK(sc);
2142 
2143 	ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
2144 }
2145 
2146 /*
2147  * bridge_enqueue:
2148  *
2149  *	Enqueue a packet on a bridge member interface.
2150  *
2151  */
2152 static int
2153 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
2154 {
2155 	int len, err = 0;
2156 	short mflags;
2157 	struct mbuf *m0;
2158 
2159 	/* We may be sending a fragment so traverse the mbuf */
2160 	for (; m; m = m0) {
2161 		m0 = m->m_nextpkt;
2162 		m->m_nextpkt = NULL;
2163 		len = m->m_pkthdr.len;
2164 		mflags = m->m_flags;
2165 
2166 		/*
2167 		 * If underlying interface can not do VLAN tag insertion itself
2168 		 * then attach a packet tag that holds it.
2169 		 */
2170 		if ((m->m_flags & M_VLANTAG) &&
2171 		    (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
2172 			m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
2173 			if (m == NULL) {
2174 				if_printf(dst_ifp,
2175 				    "unable to prepend VLAN header\n");
2176 				if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
2177 				continue;
2178 			}
2179 			m->m_flags &= ~M_VLANTAG;
2180 		}
2181 
2182 		M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
2183 		if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
2184 			int n;
2185 
2186 			for (m = m0, n = 1; m != NULL; m = m0, n++) {
2187 				m0 = m->m_nextpkt;
2188 				m_freem(m);
2189 			}
2190 			if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, n);
2191 			break;
2192 		}
2193 
2194 		if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1);
2195 		if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len);
2196 		if (mflags & M_MCAST)
2197 			if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1);
2198 	}
2199 
2200 	return (err);
2201 }
2202 
2203 /*
2204  * bridge_dummynet:
2205  *
2206  * 	Receive a queued packet from dummynet and pass it on to the output
2207  * 	interface.
2208  *
2209  *	The mbuf has the Ethernet header already attached.
2210  */
2211 static void
2212 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
2213 {
2214 	struct bridge_iflist *bif = ifp->if_bridge;
2215 	struct bridge_softc *sc = NULL;
2216 
2217 	if (bif)
2218 		sc = bif->bif_sc;
2219 
2220 	/*
2221 	 * The packet didnt originate from a member interface. This should only
2222 	 * ever happen if a member interface is removed while packets are
2223 	 * queued for it.
2224 	 */
2225 	if (sc == NULL) {
2226 		m_freem(m);
2227 		return;
2228 	}
2229 
2230 	if (PFIL_HOOKED_OUT_46) {
2231 		if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
2232 			return;
2233 		if (m == NULL)
2234 			return;
2235 	}
2236 
2237 	bridge_enqueue(sc, ifp, m);
2238 }
2239 
2240 /*
2241  * bridge_output:
2242  *
2243  *	Send output from a bridge member interface.  This
2244  *	performs the bridging function for locally originated
2245  *	packets.
2246  *
2247  *	The mbuf has the Ethernet header already attached.  We must
2248  *	enqueue or free the mbuf before returning.
2249  */
2250 static int
2251 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2252     struct rtentry *rt)
2253 {
2254 	struct ether_header *eh;
2255 	struct bridge_iflist *sbif;
2256 	struct ifnet *bifp, *dst_if;
2257 	struct bridge_softc *sc;
2258 	ether_vlanid_t vlan;
2259 
2260 	NET_EPOCH_ASSERT();
2261 
2262 	if (m->m_len < ETHER_HDR_LEN) {
2263 		m = m_pullup(m, ETHER_HDR_LEN);
2264 		if (m == NULL)
2265 			return (0);
2266 	}
2267 
2268 	sbif = ifp->if_bridge;
2269 	sc = sbif->bif_sc;
2270 	bifp = sc->sc_ifp;
2271 
2272 	eh = mtod(m, struct ether_header *);
2273 	vlan = VLANTAGOF(m);
2274 
2275 	/*
2276 	 * If bridge is down, but the original output interface is up,
2277 	 * go ahead and send out that interface.  Otherwise, the packet
2278 	 * is dropped below.
2279 	 */
2280 	if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2281 		dst_if = ifp;
2282 		goto sendunicast;
2283 	}
2284 
2285 	/*
2286 	 * If the packet is a multicast, or we don't know a better way to
2287 	 * get there, send to all interfaces.
2288 	 */
2289 	if (ETHER_IS_MULTICAST(eh->ether_dhost))
2290 		dst_if = NULL;
2291 	else
2292 		dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2293 	/* Tap any traffic not passing back out the originating interface */
2294 	if (dst_if != ifp)
2295 		ETHER_BPF_MTAP(bifp, m);
2296 	if (dst_if == NULL) {
2297 		struct bridge_iflist *bif;
2298 		struct mbuf *mc;
2299 		int used = 0;
2300 
2301 		bridge_span(sc, m);
2302 
2303 		CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2304 			dst_if = bif->bif_ifp;
2305 
2306 			if (dst_if->if_type == IFT_GIF)
2307 				continue;
2308 			if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2309 				continue;
2310 
2311 			/*
2312 			 * If this is not the original output interface,
2313 			 * and the interface is participating in spanning
2314 			 * tree, make sure the port is in a state that
2315 			 * allows forwarding.
2316 			 */
2317 			if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2318 			    bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2319 				continue;
2320 
2321 			if (CK_LIST_NEXT(bif, bif_next) == NULL) {
2322 				used = 1;
2323 				mc = m;
2324 			} else {
2325 				mc = m_dup(m, M_NOWAIT);
2326 				if (mc == NULL) {
2327 					if_inc_counter(bifp, IFCOUNTER_OERRORS, 1);
2328 					continue;
2329 				}
2330 			}
2331 
2332 			bridge_enqueue(sc, dst_if, mc);
2333 		}
2334 		if (used == 0)
2335 			m_freem(m);
2336 		return (0);
2337 	}
2338 
2339 sendunicast:
2340 	/*
2341 	 * XXX Spanning tree consideration here?
2342 	 */
2343 
2344 	bridge_span(sc, m);
2345 	if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2346 		m_freem(m);
2347 		return (0);
2348 	}
2349 
2350 	bridge_enqueue(sc, dst_if, m);
2351 	return (0);
2352 }
2353 
2354 /*
2355  * bridge_transmit:
2356  *
2357  *	Do output on a bridge.
2358  *
2359  */
2360 static int
2361 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2362 {
2363 	struct bridge_softc *sc;
2364 	struct ether_header *eh;
2365 	struct ifnet *dst_if;
2366 	int error = 0;
2367 
2368 	sc = ifp->if_softc;
2369 
2370 	ETHER_BPF_MTAP(ifp, m);
2371 
2372 	eh = mtod(m, struct ether_header *);
2373 
2374 	if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2375 	    (dst_if = bridge_rtlookup(sc, eh->ether_dhost, DOT1Q_VID_NULL)) !=
2376 	    NULL) {
2377 		error = bridge_enqueue(sc, dst_if, m);
2378 	} else
2379 		bridge_broadcast(sc, ifp, m, 0);
2380 
2381 	return (error);
2382 }
2383 
2384 #ifdef ALTQ
2385 static void
2386 bridge_altq_start(if_t ifp)
2387 {
2388 	struct ifaltq *ifq = &ifp->if_snd;
2389 	struct mbuf *m;
2390 
2391 	IFQ_LOCK(ifq);
2392 	IFQ_DEQUEUE_NOLOCK(ifq, m);
2393 	while (m != NULL) {
2394 		bridge_transmit(ifp, m);
2395 		IFQ_DEQUEUE_NOLOCK(ifq, m);
2396 	}
2397 	IFQ_UNLOCK(ifq);
2398 }
2399 
2400 static int
2401 bridge_altq_transmit(if_t ifp, struct mbuf *m)
2402 {
2403 	int err;
2404 
2405 	if (ALTQ_IS_ENABLED(&ifp->if_snd)) {
2406 		IFQ_ENQUEUE(&ifp->if_snd, m, err);
2407 		if (err == 0)
2408 			bridge_altq_start(ifp);
2409 	} else
2410 		err = bridge_transmit(ifp, m);
2411 
2412 	return (err);
2413 }
2414 #endif	/* ALTQ */
2415 
2416 /*
2417  * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2418  */
2419 static void
2420 bridge_qflush(struct ifnet *ifp __unused)
2421 {
2422 }
2423 
2424 /*
2425  * bridge_forward:
2426  *
2427  *	The forwarding function of the bridge.
2428  *
2429  *	NOTE: Releases the lock on return.
2430  */
2431 static void
2432 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2433     struct mbuf *m)
2434 {
2435 	struct bridge_iflist *dbif;
2436 	struct ifnet *src_if, *dst_if, *ifp;
2437 	struct ether_header *eh;
2438 	uint16_t vlan;
2439 	uint8_t *dst;
2440 	int error;
2441 
2442 	NET_EPOCH_ASSERT();
2443 
2444 	src_if = m->m_pkthdr.rcvif;
2445 	ifp = sc->sc_ifp;
2446 
2447 	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2448 	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2449 	vlan = VLANTAGOF(m);
2450 
2451 	if ((sbif->bif_flags & IFBIF_STP) &&
2452 	    sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2453 		goto drop;
2454 
2455 	eh = mtod(m, struct ether_header *);
2456 	dst = eh->ether_dhost;
2457 
2458 	/* If the interface is learning, record the address. */
2459 	if (sbif->bif_flags & IFBIF_LEARNING) {
2460 		error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2461 		    sbif, 0, IFBAF_DYNAMIC);
2462 		/*
2463 		 * If the interface has addresses limits then deny any source
2464 		 * that is not in the cache.
2465 		 */
2466 		if (error && sbif->bif_addrmax)
2467 			goto drop;
2468 	}
2469 
2470 	if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2471 	    sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2472 		goto drop;
2473 
2474 #ifdef DEV_NETMAP
2475 	/*
2476 	 * Hand the packet to netmap only if it wasn't injected by netmap
2477 	 * itself.
2478 	 */
2479 	if ((m->m_flags & M_BRIDGE_INJECT) == 0 &&
2480 	    (if_getcapenable(ifp) & IFCAP_NETMAP) != 0) {
2481 		ifp->if_input(ifp, m);
2482 		return;
2483 	}
2484 	m->m_flags &= ~M_BRIDGE_INJECT;
2485 #endif
2486 
2487 	/*
2488 	 * At this point, the port either doesn't participate
2489 	 * in spanning tree or it is in the forwarding state.
2490 	 */
2491 
2492 	/*
2493 	 * If the packet is unicast, destined for someone on
2494 	 * "this" side of the bridge, drop it.
2495 	 */
2496 	if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2497 		dst_if = bridge_rtlookup(sc, dst, vlan);
2498 		if (src_if == dst_if)
2499 			goto drop;
2500 	} else {
2501 		/*
2502 		 * Check if its a reserved multicast address, any address
2503 		 * listed in 802.1D section 7.12.6 may not be forwarded by the
2504 		 * bridge.
2505 		 * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2506 		 */
2507 		if (dst[0] == 0x01 && dst[1] == 0x80 &&
2508 		    dst[2] == 0xc2 && dst[3] == 0x00 &&
2509 		    dst[4] == 0x00 && dst[5] <= 0x0f)
2510 			goto drop;
2511 
2512 		/* ...forward it to all interfaces. */
2513 		if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
2514 		dst_if = NULL;
2515 	}
2516 
2517 	/*
2518 	 * If we have a destination interface which is a member of our bridge,
2519 	 * OR this is a unicast packet, push it through the bpf(4) machinery.
2520 	 * For broadcast or multicast packets, don't bother because it will
2521 	 * be reinjected into ether_input. We do this before we pass the packets
2522 	 * through the pfil(9) framework, as it is possible that pfil(9) will
2523 	 * drop the packet, or possibly modify it, making it difficult to debug
2524 	 * firewall issues on the bridge.
2525 	 */
2526 	if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2527 		ETHER_BPF_MTAP(ifp, m);
2528 
2529 	/* run the packet filter */
2530 	if (PFIL_HOOKED_IN_46) {
2531 		if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2532 			return;
2533 		if (m == NULL)
2534 			return;
2535 	}
2536 
2537 	if (dst_if == NULL) {
2538 		bridge_broadcast(sc, src_if, m, 1);
2539 		return;
2540 	}
2541 
2542 	/*
2543 	 * At this point, we're dealing with a unicast frame
2544 	 * going to a different interface.
2545 	 */
2546 	if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2547 		goto drop;
2548 
2549 	dbif = bridge_lookup_member_if(sc, dst_if);
2550 	if (dbif == NULL)
2551 		/* Not a member of the bridge (anymore?) */
2552 		goto drop;
2553 
2554 	/* Private segments can not talk to each other */
2555 	if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2556 		goto drop;
2557 
2558 	if ((dbif->bif_flags & IFBIF_STP) &&
2559 	    dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2560 		goto drop;
2561 
2562 	if (PFIL_HOOKED_OUT_46) {
2563 		if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2564 			return;
2565 		if (m == NULL)
2566 			return;
2567 	}
2568 
2569 	bridge_enqueue(sc, dst_if, m);
2570 	return;
2571 
2572 drop:
2573 	m_freem(m);
2574 }
2575 
2576 /*
2577  * bridge_input:
2578  *
2579  *	Receive input from a member interface.  Queue the packet for
2580  *	bridging if it is not for us.
2581  */
2582 static struct mbuf *
2583 bridge_input(struct ifnet *ifp, struct mbuf *m)
2584 {
2585 	struct bridge_softc *sc = NULL;
2586 	struct bridge_iflist *bif, *bif2;
2587 	struct ifnet *bifp;
2588 	struct ether_header *eh;
2589 	struct mbuf *mc, *mc2;
2590 	ether_vlanid_t vlan;
2591 	int error;
2592 
2593 	NET_EPOCH_ASSERT();
2594 
2595 	eh = mtod(m, struct ether_header *);
2596 	vlan = VLANTAGOF(m);
2597 
2598 	bif = ifp->if_bridge;
2599 	if (bif)
2600 		sc = bif->bif_sc;
2601 
2602 	if (sc == NULL) {
2603 		/*
2604 		 * This packet originated from the bridge itself, so it must
2605 		 * have been transmitted by netmap.  Derive the "source"
2606 		 * interface from the source address and drop the packet if the
2607 		 * source address isn't known.
2608 		 */
2609 		KASSERT((m->m_flags & M_BRIDGE_INJECT) != 0,
2610 		    ("%s: ifnet %p missing a bridge softc", __func__, ifp));
2611 		sc = if_getsoftc(ifp);
2612 		ifp = bridge_rtlookup(sc, eh->ether_shost, vlan);
2613 		if (ifp == NULL) {
2614 			if_inc_counter(sc->sc_ifp, IFCOUNTER_IERRORS, 1);
2615 			m_freem(m);
2616 			return (NULL);
2617 		}
2618 		m->m_pkthdr.rcvif = ifp;
2619 	}
2620 	bifp = sc->sc_ifp;
2621 	if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2622 		return (m);
2623 
2624 	/*
2625 	 * Implement support for bridge monitoring. If this flag has been
2626 	 * set on this interface, discard the packet once we push it through
2627 	 * the bpf(4) machinery, but before we do, increment the byte and
2628 	 * packet counters associated with this interface.
2629 	 */
2630 	if ((bifp->if_flags & IFF_MONITOR) != 0) {
2631 		m->m_pkthdr.rcvif  = bifp;
2632 		ETHER_BPF_MTAP(bifp, m);
2633 		if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);
2634 		if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2635 		m_freem(m);
2636 		return (NULL);
2637 	}
2638 
2639 	bridge_span(sc, m);
2640 
2641 	if (m->m_flags & (M_BCAST|M_MCAST)) {
2642 		/* Tap off 802.1D packets; they do not get forwarded. */
2643 		if (memcmp(eh->ether_dhost, bstp_etheraddr,
2644 		    ETHER_ADDR_LEN) == 0) {
2645 			bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2646 			return (NULL);
2647 		}
2648 
2649 		if ((bif->bif_flags & IFBIF_STP) &&
2650 		    bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2651 			return (m);
2652 		}
2653 
2654 		/*
2655 		 * Make a deep copy of the packet and enqueue the copy
2656 		 * for bridge processing; return the original packet for
2657 		 * local processing.
2658 		 */
2659 		mc = m_dup(m, M_NOWAIT);
2660 		if (mc == NULL) {
2661 			return (m);
2662 		}
2663 
2664 		/* Perform the bridge forwarding function with the copy. */
2665 		bridge_forward(sc, bif, mc);
2666 
2667 #ifdef DEV_NETMAP
2668 		/*
2669 		 * If netmap is enabled and has not already seen this packet,
2670 		 * then it will be consumed by bridge_forward().
2671 		 */
2672 		if ((if_getcapenable(bifp) & IFCAP_NETMAP) != 0 &&
2673 		    (m->m_flags & M_BRIDGE_INJECT) == 0) {
2674 			m_freem(m);
2675 			return (NULL);
2676 		}
2677 #endif
2678 
2679 		/*
2680 		 * Reinject the mbuf as arriving on the bridge so we have a
2681 		 * chance at claiming multicast packets. We can not loop back
2682 		 * here from ether_input as a bridge is never a member of a
2683 		 * bridge.
2684 		 */
2685 		KASSERT(bifp->if_bridge == NULL,
2686 		    ("loop created in bridge_input"));
2687 		mc2 = m_dup(m, M_NOWAIT);
2688 		if (mc2 != NULL) {
2689 			/* Keep the layer3 header aligned */
2690 			int i = min(mc2->m_pkthdr.len, max_protohdr);
2691 			mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2692 		}
2693 		if (mc2 != NULL) {
2694 			mc2->m_pkthdr.rcvif = bifp;
2695 			mc2->m_flags &= ~M_BRIDGE_INJECT;
2696 			sc->sc_if_input(bifp, mc2);
2697 		}
2698 
2699 		/* Return the original packet for local processing. */
2700 		return (m);
2701 	}
2702 
2703 	if ((bif->bif_flags & IFBIF_STP) &&
2704 	    bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2705 		return (m);
2706 	}
2707 
2708 #if defined(INET) || defined(INET6)
2709 #define	CARP_CHECK_WE_ARE_DST(iface) \
2710 	((iface)->if_carp && (*carp_forus_p)((iface), eh->ether_dhost))
2711 #define	CARP_CHECK_WE_ARE_SRC(iface) \
2712 	((iface)->if_carp && (*carp_forus_p)((iface), eh->ether_shost))
2713 #else
2714 #define	CARP_CHECK_WE_ARE_DST(iface)	false
2715 #define	CARP_CHECK_WE_ARE_SRC(iface)	false
2716 #endif
2717 
2718 #ifdef DEV_NETMAP
2719 #define	GRAB_FOR_NETMAP(ifp, m) do {					\
2720 	if ((if_getcapenable(ifp) & IFCAP_NETMAP) != 0 &&		\
2721 	    ((m)->m_flags & M_BRIDGE_INJECT) == 0) {			\
2722 		(ifp)->if_input(ifp, m);				\
2723 		return (NULL);						\
2724 	}								\
2725 } while (0)
2726 #else
2727 #define	GRAB_FOR_NETMAP(ifp, m)
2728 #endif
2729 
2730 #define GRAB_OUR_PACKETS(iface)						\
2731 	if ((iface)->if_type == IFT_GIF)				\
2732 		continue;						\
2733 	/* It is destined for us. */					\
2734 	if (memcmp(IF_LLADDR(iface), eh->ether_dhost, ETHER_ADDR_LEN) == 0 || \
2735 	    CARP_CHECK_WE_ARE_DST(iface)) {				\
2736 		if (bif->bif_flags & IFBIF_LEARNING) {			\
2737 			error = bridge_rtupdate(sc, eh->ether_shost,	\
2738 			    vlan, bif, 0, IFBAF_DYNAMIC);		\
2739 			if (error && bif->bif_addrmax) {		\
2740 				m_freem(m);				\
2741 				return (NULL);				\
2742 			}						\
2743 		}							\
2744 		m->m_pkthdr.rcvif = iface;				\
2745 		if ((iface) == ifp) {					\
2746 			/* Skip bridge processing... src == dest */	\
2747 			return (m);					\
2748 		}							\
2749 		/* It's passing over or to the bridge, locally. */	\
2750 		ETHER_BPF_MTAP(bifp, m);				\
2751 		if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);		\
2752 		if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);\
2753 		/* Hand the packet over to netmap if necessary. */	\
2754 		GRAB_FOR_NETMAP(bifp, m);				\
2755 		/* Filter on the physical interface. */			\
2756 		if (V_pfil_local_phys && PFIL_HOOKED_IN_46) {		\
2757 			if (bridge_pfil(&m, NULL, ifp,			\
2758 			    PFIL_IN) != 0 || m == NULL) {		\
2759 				return (NULL);				\
2760 			}						\
2761 		}							\
2762 		if ((iface) != bifp)					\
2763 			ETHER_BPF_MTAP(iface, m);			\
2764 		return (m);						\
2765 	}								\
2766 									\
2767 	/* We just received a packet that we sent out. */		\
2768 	if (memcmp(IF_LLADDR(iface), eh->ether_shost, ETHER_ADDR_LEN) == 0 || \
2769 	    CARP_CHECK_WE_ARE_SRC(iface)) {				\
2770 		m_freem(m);						\
2771 		return (NULL);						\
2772 	}
2773 
2774 	/*
2775 	 * Unicast.  Make sure it's not for the bridge.
2776 	 */
2777 	do { GRAB_OUR_PACKETS(bifp) } while (0);
2778 
2779 	/*
2780 	 * We only need to check members interfaces if member_ifaddrs is
2781 	 * enabled; otherwise we should have never traffic destined for a
2782 	 * member's lladdr.
2783 	 */
2784 
2785 	if (V_member_ifaddrs) {
2786 		/*
2787 		 * Give a chance for ifp at first priority. This will help when
2788 		 * the packet comes through the interface like VLAN's with the
2789 		 * same MACs on several interfaces from the same bridge. This
2790 		 * also will save some CPU cycles in case the destination
2791 		 * interface and the input interface (eq ifp) are the same.
2792 		 */
2793 		do { GRAB_OUR_PACKETS(ifp) } while (0);
2794 
2795 		/* Now check the all bridge members. */
2796 		CK_LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2797 			GRAB_OUR_PACKETS(bif2->bif_ifp)
2798 		}
2799 	}
2800 
2801 #undef CARP_CHECK_WE_ARE_DST
2802 #undef CARP_CHECK_WE_ARE_SRC
2803 #undef GRAB_FOR_NETMAP
2804 #undef GRAB_OUR_PACKETS
2805 
2806 	/* Perform the bridge forwarding function. */
2807 	bridge_forward(sc, bif, m);
2808 
2809 	return (NULL);
2810 }
2811 
2812 /*
2813  * Inject a packet back into the host ethernet stack.  This will generally only
2814  * be used by netmap when an application writes to the host TX ring.  The
2815  * M_BRIDGE_INJECT flag ensures that the packet is re-routed to the bridge
2816  * interface after ethernet processing.
2817  */
2818 static void
2819 bridge_inject(struct ifnet *ifp, struct mbuf *m)
2820 {
2821 	struct bridge_softc *sc;
2822 
2823 	KASSERT((if_getcapenable(ifp) & IFCAP_NETMAP) != 0,
2824 	    ("%s: iface %s is not running in netmap mode",
2825 	    __func__, if_name(ifp)));
2826 	KASSERT((m->m_flags & M_BRIDGE_INJECT) == 0,
2827 	    ("%s: mbuf %p has M_BRIDGE_INJECT set", __func__, m));
2828 
2829 	m->m_flags |= M_BRIDGE_INJECT;
2830 	sc = if_getsoftc(ifp);
2831 	sc->sc_if_input(ifp, m);
2832 }
2833 
2834 /*
2835  * bridge_broadcast:
2836  *
2837  *	Send a frame to all interfaces that are members of
2838  *	the bridge, except for the one on which the packet
2839  *	arrived.
2840  *
2841  *	NOTE: Releases the lock on return.
2842  */
2843 static void
2844 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2845     struct mbuf *m, int runfilt)
2846 {
2847 	struct bridge_iflist *dbif, *sbif;
2848 	struct mbuf *mc;
2849 	struct ifnet *dst_if;
2850 	int used = 0, i;
2851 
2852 	NET_EPOCH_ASSERT();
2853 
2854 	sbif = bridge_lookup_member_if(sc, src_if);
2855 
2856 	/* Filter on the bridge interface before broadcasting */
2857 	if (runfilt && PFIL_HOOKED_OUT_46) {
2858 		if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2859 			return;
2860 		if (m == NULL)
2861 			return;
2862 	}
2863 
2864 	CK_LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2865 		dst_if = dbif->bif_ifp;
2866 		if (dst_if == src_if)
2867 			continue;
2868 
2869 		/* Private segments can not talk to each other */
2870 		if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2871 			continue;
2872 
2873 		if ((dbif->bif_flags & IFBIF_STP) &&
2874 		    dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2875 			continue;
2876 
2877 		if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2878 		    (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2879 			continue;
2880 
2881 		if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2882 			continue;
2883 
2884 		if (CK_LIST_NEXT(dbif, bif_next) == NULL) {
2885 			mc = m;
2886 			used = 1;
2887 		} else {
2888 			mc = m_dup(m, M_NOWAIT);
2889 			if (mc == NULL) {
2890 				if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2891 				continue;
2892 			}
2893 		}
2894 
2895 		/*
2896 		 * Filter on the output interface. Pass a NULL bridge interface
2897 		 * pointer so we do not redundantly filter on the bridge for
2898 		 * each interface we broadcast on.
2899 		 */
2900 		if (runfilt && PFIL_HOOKED_OUT_46) {
2901 			if (used == 0) {
2902 				/* Keep the layer3 header aligned */
2903 				i = min(mc->m_pkthdr.len, max_protohdr);
2904 				mc = m_copyup(mc, i, ETHER_ALIGN);
2905 				if (mc == NULL) {
2906 					if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2907 					continue;
2908 				}
2909 			}
2910 			if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2911 				continue;
2912 			if (mc == NULL)
2913 				continue;
2914 		}
2915 
2916 		bridge_enqueue(sc, dst_if, mc);
2917 	}
2918 	if (used == 0)
2919 		m_freem(m);
2920 }
2921 
2922 /*
2923  * bridge_span:
2924  *
2925  *	Duplicate a packet out one or more interfaces that are in span mode,
2926  *	the original mbuf is unmodified.
2927  */
2928 static void
2929 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2930 {
2931 	struct bridge_iflist *bif;
2932 	struct ifnet *dst_if;
2933 	struct mbuf *mc;
2934 
2935 	NET_EPOCH_ASSERT();
2936 
2937 	if (CK_LIST_EMPTY(&sc->sc_spanlist))
2938 		return;
2939 
2940 	CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2941 		dst_if = bif->bif_ifp;
2942 
2943 		if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2944 			continue;
2945 
2946 		mc = m_dup(m, M_NOWAIT);
2947 		if (mc == NULL) {
2948 			if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2949 			continue;
2950 		}
2951 
2952 		bridge_enqueue(sc, dst_if, mc);
2953 	}
2954 }
2955 
2956 /*
2957  * bridge_rtupdate:
2958  *
2959  *	Add a bridge routing entry.
2960  */
2961 static int
2962 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
2963 		ether_vlanid_t vlan, struct bridge_iflist *bif,
2964 		int setflags, uint8_t flags)
2965 {
2966 	struct bridge_rtnode *brt;
2967 	struct bridge_iflist *obif;
2968 	int error;
2969 
2970 	BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
2971 
2972 	/* Check the source address is valid and not multicast. */
2973 	if (ETHER_IS_MULTICAST(dst) ||
2974 	    (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2975 	     dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2976 		return (EINVAL);
2977 
2978 	/*
2979 	 * A route for this destination might already exist.  If so,
2980 	 * update it, otherwise create a new one.
2981 	 */
2982 	if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2983 		BRIDGE_RT_LOCK(sc);
2984 
2985 		/* Check again, now that we have the lock. There could have
2986 		 * been a race and we only want to insert this once. */
2987 		if (bridge_rtnode_lookup(sc, dst, vlan) != NULL) {
2988 			BRIDGE_RT_UNLOCK(sc);
2989 			return (0);
2990 		}
2991 
2992 		if (sc->sc_brtcnt >= sc->sc_brtmax) {
2993 			sc->sc_brtexceeded++;
2994 			BRIDGE_RT_UNLOCK(sc);
2995 			return (ENOSPC);
2996 		}
2997 		/* Check per interface address limits (if enabled) */
2998 		if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2999 			bif->bif_addrexceeded++;
3000 			BRIDGE_RT_UNLOCK(sc);
3001 			return (ENOSPC);
3002 		}
3003 
3004 		/*
3005 		 * Allocate a new bridge forwarding node, and
3006 		 * initialize the expiration time and Ethernet
3007 		 * address.
3008 		 */
3009 		brt = uma_zalloc(V_bridge_rtnode_zone, M_NOWAIT | M_ZERO);
3010 		if (brt == NULL) {
3011 			BRIDGE_RT_UNLOCK(sc);
3012 			return (ENOMEM);
3013 		}
3014 		brt->brt_vnet = curvnet;
3015 
3016 		if (bif->bif_flags & IFBIF_STICKY)
3017 			brt->brt_flags = IFBAF_STICKY;
3018 		else
3019 			brt->brt_flags = IFBAF_DYNAMIC;
3020 
3021 		memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
3022 		brt->brt_vlan = vlan;
3023 
3024 		brt->brt_dst = bif;
3025 		if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
3026 			uma_zfree(V_bridge_rtnode_zone, brt);
3027 			BRIDGE_RT_UNLOCK(sc);
3028 			return (error);
3029 		}
3030 		bif->bif_addrcnt++;
3031 
3032 		BRIDGE_RT_UNLOCK(sc);
3033 	}
3034 
3035 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
3036 	    (obif = brt->brt_dst) != bif) {
3037 		MPASS(obif != NULL);
3038 
3039 		BRIDGE_RT_LOCK(sc);
3040 		brt->brt_dst->bif_addrcnt--;
3041 		brt->brt_dst = bif;
3042 		brt->brt_dst->bif_addrcnt++;
3043 		BRIDGE_RT_UNLOCK(sc);
3044 
3045 		if (V_log_mac_flap &&
3046 		    ppsratecheck(&V_log_last, &V_log_count, V_log_interval)) {
3047 			log(LOG_NOTICE,
3048 			    "%s: mac address %6D vlan %d moved from %s to %s\n",
3049 			    sc->sc_ifp->if_xname,
3050 			    &brt->brt_addr[0], ":",
3051 			    brt->brt_vlan,
3052 			    obif->bif_ifp->if_xname,
3053 			    bif->bif_ifp->if_xname);
3054 		}
3055 	}
3056 
3057 	if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3058 		brt->brt_expire = time_uptime + sc->sc_brttimeout;
3059 	if (setflags)
3060 		brt->brt_flags = flags;
3061 
3062 	return (0);
3063 }
3064 
3065 /*
3066  * bridge_rtlookup:
3067  *
3068  *	Lookup the destination interface for an address.
3069  */
3070 static struct ifnet *
3071 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr,
3072 		ether_vlanid_t vlan)
3073 {
3074 	struct bridge_rtnode *brt;
3075 
3076 	NET_EPOCH_ASSERT();
3077 
3078 	if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
3079 		return (NULL);
3080 
3081 	return (brt->brt_ifp);
3082 }
3083 
3084 /*
3085  * bridge_rttrim:
3086  *
3087  *	Trim the routine table so that we have a number
3088  *	of routing entries less than or equal to the
3089  *	maximum number.
3090  */
3091 static void
3092 bridge_rttrim(struct bridge_softc *sc)
3093 {
3094 	struct bridge_rtnode *brt, *nbrt;
3095 
3096 	NET_EPOCH_ASSERT();
3097 	BRIDGE_RT_LOCK_ASSERT(sc);
3098 
3099 	/* Make sure we actually need to do this. */
3100 	if (sc->sc_brtcnt <= sc->sc_brtmax)
3101 		return;
3102 
3103 	/* Force an aging cycle; this might trim enough addresses. */
3104 	bridge_rtage(sc);
3105 	if (sc->sc_brtcnt <= sc->sc_brtmax)
3106 		return;
3107 
3108 	CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3109 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
3110 			bridge_rtnode_destroy(sc, brt);
3111 			if (sc->sc_brtcnt <= sc->sc_brtmax)
3112 				return;
3113 		}
3114 	}
3115 }
3116 
3117 /*
3118  * bridge_timer:
3119  *
3120  *	Aging timer for the bridge.
3121  */
3122 static void
3123 bridge_timer(void *arg)
3124 {
3125 	struct bridge_softc *sc = arg;
3126 
3127 	BRIDGE_RT_LOCK_ASSERT(sc);
3128 
3129 	/* Destruction of rtnodes requires a proper vnet context */
3130 	CURVNET_SET(sc->sc_ifp->if_vnet);
3131 	bridge_rtage(sc);
3132 
3133 	if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
3134 		callout_reset(&sc->sc_brcallout,
3135 		    bridge_rtable_prune_period * hz, bridge_timer, sc);
3136 	CURVNET_RESTORE();
3137 }
3138 
3139 /*
3140  * bridge_rtage:
3141  *
3142  *	Perform an aging cycle.
3143  */
3144 static void
3145 bridge_rtage(struct bridge_softc *sc)
3146 {
3147 	struct bridge_rtnode *brt, *nbrt;
3148 
3149 	BRIDGE_RT_LOCK_ASSERT(sc);
3150 
3151 	CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3152 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
3153 			if (time_uptime >= brt->brt_expire)
3154 				bridge_rtnode_destroy(sc, brt);
3155 		}
3156 	}
3157 }
3158 
3159 /*
3160  * bridge_rtflush:
3161  *
3162  *	Remove all dynamic addresses from the bridge.
3163  */
3164 static void
3165 bridge_rtflush(struct bridge_softc *sc, int full)
3166 {
3167 	struct bridge_rtnode *brt, *nbrt;
3168 
3169 	BRIDGE_RT_LOCK_ASSERT(sc);
3170 
3171 	CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3172 		if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3173 			bridge_rtnode_destroy(sc, brt);
3174 	}
3175 }
3176 
3177 /*
3178  * bridge_rtdaddr:
3179  *
3180  *	Remove an address from the table.
3181  */
3182 static int
3183 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr,
3184 	       ether_vlanid_t vlan)
3185 {
3186 	struct bridge_rtnode *brt;
3187 	int found = 0;
3188 
3189 	BRIDGE_RT_LOCK(sc);
3190 
3191 	/*
3192 	 * If vlan is DOT1Q_VID_RSVD_IMPL then we want to delete for all vlans
3193 	 * so the lookup may return more than one.
3194 	 */
3195 	while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
3196 		bridge_rtnode_destroy(sc, brt);
3197 		found = 1;
3198 	}
3199 
3200 	BRIDGE_RT_UNLOCK(sc);
3201 
3202 	return (found ? 0 : ENOENT);
3203 }
3204 
3205 /*
3206  * bridge_rtdelete:
3207  *
3208  *	Delete routes to a speicifc member interface.
3209  */
3210 static void
3211 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
3212 {
3213 	struct bridge_rtnode *brt, *nbrt;
3214 
3215 	BRIDGE_RT_LOCK_ASSERT(sc);
3216 
3217 	CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3218 		if (brt->brt_ifp == ifp && (full ||
3219 			    (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
3220 			bridge_rtnode_destroy(sc, brt);
3221 	}
3222 }
3223 
3224 /*
3225  * bridge_rtable_init:
3226  *
3227  *	Initialize the route table for this bridge.
3228  */
3229 static void
3230 bridge_rtable_init(struct bridge_softc *sc)
3231 {
3232 	int i;
3233 
3234 	sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
3235 	    M_DEVBUF, M_WAITOK);
3236 
3237 	for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
3238 		CK_LIST_INIT(&sc->sc_rthash[i]);
3239 
3240 	sc->sc_rthash_key = arc4random();
3241 	CK_LIST_INIT(&sc->sc_rtlist);
3242 }
3243 
3244 /*
3245  * bridge_rtable_fini:
3246  *
3247  *	Deconstruct the route table for this bridge.
3248  */
3249 static void
3250 bridge_rtable_fini(struct bridge_softc *sc)
3251 {
3252 
3253 	KASSERT(sc->sc_brtcnt == 0,
3254 	    ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
3255 	free(sc->sc_rthash, M_DEVBUF);
3256 }
3257 
3258 /*
3259  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3260  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3261  */
3262 #define	mix(a, b, c)							\
3263 do {									\
3264 	a -= b; a -= c; a ^= (c >> 13);					\
3265 	b -= c; b -= a; b ^= (a << 8);					\
3266 	c -= a; c -= b; c ^= (b >> 13);					\
3267 	a -= b; a -= c; a ^= (c >> 12);					\
3268 	b -= c; b -= a; b ^= (a << 16);					\
3269 	c -= a; c -= b; c ^= (b >> 5);					\
3270 	a -= b; a -= c; a ^= (c >> 3);					\
3271 	b -= c; b -= a; b ^= (a << 10);					\
3272 	c -= a; c -= b; c ^= (b >> 15);					\
3273 } while (/*CONSTCOND*/0)
3274 
3275 static __inline uint32_t
3276 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
3277 {
3278 	uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
3279 
3280 	b += addr[5] << 8;
3281 	b += addr[4];
3282 	a += addr[3] << 24;
3283 	a += addr[2] << 16;
3284 	a += addr[1] << 8;
3285 	a += addr[0];
3286 
3287 	mix(a, b, c);
3288 
3289 	return (c & BRIDGE_RTHASH_MASK);
3290 }
3291 
3292 #undef mix
3293 
3294 static int
3295 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
3296 {
3297 	int i, d;
3298 
3299 	for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
3300 		d = ((int)a[i]) - ((int)b[i]);
3301 	}
3302 
3303 	return (d);
3304 }
3305 
3306 /*
3307  * bridge_rtnode_lookup:
3308  *
3309  *	Look up a bridge route node for the specified destination. Compare the
3310  *	vlan id or if zero then just return the first match.
3311  */
3312 static struct bridge_rtnode *
3313 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr,
3314 		     ether_vlanid_t vlan)
3315 {
3316 	struct bridge_rtnode *brt;
3317 	uint32_t hash;
3318 	int dir;
3319 
3320 	BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(sc);
3321 
3322 	hash = bridge_rthash(sc, addr);
3323 	CK_LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
3324 		dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3325 		if (dir == 0 && (brt->brt_vlan == vlan || vlan == DOT1Q_VID_RSVD_IMPL))
3326 			return (brt);
3327 		if (dir > 0)
3328 			return (NULL);
3329 	}
3330 
3331 	return (NULL);
3332 }
3333 
3334 /*
3335  * bridge_rtnode_insert:
3336  *
3337  *	Insert the specified bridge node into the route table.  We
3338  *	assume the entry is not already in the table.
3339  */
3340 static int
3341 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3342 {
3343 	struct bridge_rtnode *lbrt;
3344 	uint32_t hash;
3345 	int dir;
3346 
3347 	BRIDGE_RT_LOCK_ASSERT(sc);
3348 
3349 	hash = bridge_rthash(sc, brt->brt_addr);
3350 
3351 	lbrt = CK_LIST_FIRST(&sc->sc_rthash[hash]);
3352 	if (lbrt == NULL) {
3353 		CK_LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
3354 		goto out;
3355 	}
3356 
3357 	do {
3358 		dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3359 		if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
3360 			return (EEXIST);
3361 		if (dir > 0) {
3362 			CK_LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3363 			goto out;
3364 		}
3365 		if (CK_LIST_NEXT(lbrt, brt_hash) == NULL) {
3366 			CK_LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3367 			goto out;
3368 		}
3369 		lbrt = CK_LIST_NEXT(lbrt, brt_hash);
3370 	} while (lbrt != NULL);
3371 
3372 #ifdef DIAGNOSTIC
3373 	panic("bridge_rtnode_insert: impossible");
3374 #endif
3375 
3376 out:
3377 	CK_LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
3378 	sc->sc_brtcnt++;
3379 
3380 	return (0);
3381 }
3382 
3383 static void
3384 bridge_rtnode_destroy_cb(struct epoch_context *ctx)
3385 {
3386 	struct bridge_rtnode *brt;
3387 
3388 	brt = __containerof(ctx, struct bridge_rtnode, brt_epoch_ctx);
3389 
3390 	CURVNET_SET(brt->brt_vnet);
3391 	uma_zfree(V_bridge_rtnode_zone, brt);
3392 	CURVNET_RESTORE();
3393 }
3394 
3395 /*
3396  * bridge_rtnode_destroy:
3397  *
3398  *	Destroy a bridge rtnode.
3399  */
3400 static void
3401 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3402 {
3403 	BRIDGE_RT_LOCK_ASSERT(sc);
3404 
3405 	CK_LIST_REMOVE(brt, brt_hash);
3406 
3407 	CK_LIST_REMOVE(brt, brt_list);
3408 	sc->sc_brtcnt--;
3409 	brt->brt_dst->bif_addrcnt--;
3410 
3411 	NET_EPOCH_CALL(bridge_rtnode_destroy_cb, &brt->brt_epoch_ctx);
3412 }
3413 
3414 /*
3415  * bridge_rtable_expire:
3416  *
3417  *	Set the expiry time for all routes on an interface.
3418  */
3419 static void
3420 bridge_rtable_expire(struct ifnet *ifp, int age)
3421 {
3422 	struct bridge_iflist *bif = NULL;
3423 	struct bridge_softc *sc = NULL;
3424 	struct bridge_rtnode *brt;
3425 
3426 	CURVNET_SET(ifp->if_vnet);
3427 
3428 	bif = ifp->if_bridge;
3429 	if (bif)
3430 		sc = bif->bif_sc;
3431 	MPASS(sc != NULL);
3432 	BRIDGE_RT_LOCK(sc);
3433 
3434 	/*
3435 	 * If the age is zero then flush, otherwise set all the expiry times to
3436 	 * age for the interface
3437 	 */
3438 	if (age == 0)
3439 		bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3440 	else {
3441 		CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3442 			/* Cap the expiry time to 'age' */
3443 			if (brt->brt_ifp == ifp &&
3444 			    brt->brt_expire > time_uptime + age &&
3445 			    (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3446 				brt->brt_expire = time_uptime + age;
3447 		}
3448 	}
3449 	BRIDGE_RT_UNLOCK(sc);
3450 	CURVNET_RESTORE();
3451 }
3452 
3453 /*
3454  * bridge_state_change:
3455  *
3456  *	Callback from the bridgestp code when a port changes states.
3457  */
3458 static void
3459 bridge_state_change(struct ifnet *ifp, int state)
3460 {
3461 	struct bridge_iflist *bif = ifp->if_bridge;
3462 	struct bridge_softc *sc = bif->bif_sc;
3463 	static const char *stpstates[] = {
3464 		"disabled",
3465 		"listening",
3466 		"learning",
3467 		"forwarding",
3468 		"blocking",
3469 		"discarding"
3470 	};
3471 
3472 	CURVNET_SET(ifp->if_vnet);
3473 	if (V_log_stp)
3474 		log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3475 		    sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3476 	CURVNET_RESTORE();
3477 }
3478 
3479 /*
3480  * Send bridge packets through pfil if they are one of the types pfil can deal
3481  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3482  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3483  * that interface.
3484  */
3485 static int
3486 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3487 {
3488 	int snap, error, i;
3489 	struct ether_header *eh1, eh2;
3490 	struct llc llc1;
3491 	u_int16_t ether_type;
3492 	pfil_return_t rv;
3493 #ifdef INET
3494 	struct ip *ip = NULL;
3495 	int hlen = 0;
3496 #endif
3497 
3498 	snap = 0;
3499 	error = -1;	/* Default error if not error == 0 */
3500 
3501 #if 0
3502 	/* we may return with the IP fields swapped, ensure its not shared */
3503 	KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3504 #endif
3505 
3506 	if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3507 		return (0); /* filtering is disabled */
3508 
3509 	i = min((*mp)->m_pkthdr.len, max_protohdr);
3510 	if ((*mp)->m_len < i) {
3511 	    *mp = m_pullup(*mp, i);
3512 	    if (*mp == NULL) {
3513 		printf("%s: m_pullup failed\n", __func__);
3514 		return (-1);
3515 	    }
3516 	}
3517 
3518 	eh1 = mtod(*mp, struct ether_header *);
3519 	ether_type = ntohs(eh1->ether_type);
3520 
3521 	/*
3522 	 * Check for SNAP/LLC.
3523 	 */
3524 	if (ether_type < ETHERMTU) {
3525 		struct llc *llc2 = (struct llc *)(eh1 + 1);
3526 
3527 		if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3528 		    llc2->llc_dsap == LLC_SNAP_LSAP &&
3529 		    llc2->llc_ssap == LLC_SNAP_LSAP &&
3530 		    llc2->llc_control == LLC_UI) {
3531 			ether_type = htons(llc2->llc_un.type_snap.ether_type);
3532 			snap = 1;
3533 		}
3534 	}
3535 
3536 	/*
3537 	 * If we're trying to filter bridge traffic, only look at traffic for
3538 	 * protocols available in the kernel (IPv4 and/or IPv6) to avoid
3539 	 * passing traffic for an unsupported protocol to the filter.  This is
3540 	 * lame since if we really wanted, say, an AppleTalk filter, we are
3541 	 * hosed, but of course we don't have an AppleTalk filter to begin
3542 	 * with.  (Note that since pfil doesn't understand ARP it will pass
3543 	 * *ALL* ARP traffic.)
3544 	 */
3545 	switch (ether_type) {
3546 #ifdef INET
3547 		case ETHERTYPE_ARP:
3548 		case ETHERTYPE_REVARP:
3549 			if (V_pfil_ipfw_arp == 0)
3550 				return (0); /* Automatically pass */
3551 
3552 			/* FALLTHROUGH */
3553 		case ETHERTYPE_IP:
3554 #endif
3555 #ifdef INET6
3556 		case ETHERTYPE_IPV6:
3557 #endif /* INET6 */
3558 			break;
3559 
3560 		default:
3561 			/*
3562 			 * We get here if the packet isn't from a supported
3563 			 * protocol.  Check to see if the user wants to pass
3564 			 * non-IP packets, these will not be checked by pfil(9)
3565 			 * and passed unconditionally so the default is to
3566 			 * drop.
3567 			 */
3568 			if (V_pfil_onlyip)
3569 				goto bad;
3570 	}
3571 
3572 	/* Run the packet through pfil before stripping link headers */
3573 	if (PFIL_HOOKED_OUT(V_link_pfil_head) && V_pfil_ipfw != 0 &&
3574 	    dir == PFIL_OUT && ifp != NULL) {
3575 		switch (pfil_mbuf_out(V_link_pfil_head, mp, ifp, NULL)) {
3576 		case PFIL_DROPPED:
3577 			return (EACCES);
3578 		case PFIL_CONSUMED:
3579 			return (0);
3580 		}
3581 	}
3582 
3583 	/* Strip off the Ethernet header and keep a copy. */
3584 	m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3585 	m_adj(*mp, ETHER_HDR_LEN);
3586 
3587 	/* Strip off snap header, if present */
3588 	if (snap) {
3589 		m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3590 		m_adj(*mp, sizeof(struct llc));
3591 	}
3592 
3593 	/*
3594 	 * Check the IP header for alignment and errors
3595 	 */
3596 	if (dir == PFIL_IN) {
3597 		switch (ether_type) {
3598 #ifdef INET
3599 			case ETHERTYPE_IP:
3600 				error = bridge_ip_checkbasic(mp);
3601 				break;
3602 #endif
3603 #ifdef INET6
3604 			case ETHERTYPE_IPV6:
3605 				error = bridge_ip6_checkbasic(mp);
3606 				break;
3607 #endif /* INET6 */
3608 			default:
3609 				error = 0;
3610 		}
3611 		if (error)
3612 			goto bad;
3613 	}
3614 
3615 	error = 0;
3616 
3617 	/*
3618 	 * Run the packet through pfil
3619 	 */
3620 	rv = PFIL_PASS;
3621 	switch (ether_type) {
3622 #ifdef INET
3623 	case ETHERTYPE_IP:
3624 		/*
3625 		 * Run pfil on the member interface and the bridge, both can
3626 		 * be skipped by clearing pfil_member or pfil_bridge.
3627 		 *
3628 		 * Keep the order:
3629 		 *   in_if -> bridge_if -> out_if
3630 		 */
3631 		if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3632 		    pfil_mbuf_out(V_inet_pfil_head, mp, bifp, NULL)) !=
3633 		    PFIL_PASS)
3634 			break;
3635 
3636 		if (V_pfil_member && ifp != NULL) {
3637 			rv = (dir == PFIL_OUT) ?
3638 			    pfil_mbuf_out(V_inet_pfil_head, mp, ifp, NULL) :
3639 			    pfil_mbuf_in(V_inet_pfil_head, mp, ifp, NULL);
3640 			if (rv != PFIL_PASS)
3641 				break;
3642 		}
3643 
3644 		if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3645 		    pfil_mbuf_in(V_inet_pfil_head, mp, bifp, NULL)) !=
3646 		    PFIL_PASS)
3647 			break;
3648 
3649 		/* check if we need to fragment the packet */
3650 		/* bridge_fragment generates a mbuf chain of packets */
3651 		/* that already include eth headers */
3652 		if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) {
3653 			i = (*mp)->m_pkthdr.len;
3654 			if (i > ifp->if_mtu) {
3655 				error = bridge_fragment(ifp, mp, &eh2, snap,
3656 					    &llc1);
3657 				return (error);
3658 			}
3659 		}
3660 
3661 		/* Recalculate the ip checksum. */
3662 		ip = mtod(*mp, struct ip *);
3663 		hlen = ip->ip_hl << 2;
3664 		if (hlen < sizeof(struct ip))
3665 			goto bad;
3666 		if (hlen > (*mp)->m_len) {
3667 			if ((*mp = m_pullup(*mp, hlen)) == NULL)
3668 				goto bad;
3669 			ip = mtod(*mp, struct ip *);
3670 			if (ip == NULL)
3671 				goto bad;
3672 		}
3673 		ip->ip_sum = 0;
3674 		if (hlen == sizeof(struct ip))
3675 			ip->ip_sum = in_cksum_hdr(ip);
3676 		else
3677 			ip->ip_sum = in_cksum(*mp, hlen);
3678 
3679 		break;
3680 #endif /* INET */
3681 #ifdef INET6
3682 	case ETHERTYPE_IPV6:
3683 		if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3684 		    pfil_mbuf_out(V_inet6_pfil_head, mp, bifp, NULL)) !=
3685 		    PFIL_PASS)
3686 			break;
3687 
3688 		if (V_pfil_member && ifp != NULL) {
3689 			rv = (dir == PFIL_OUT) ?
3690 			    pfil_mbuf_out(V_inet6_pfil_head, mp, ifp, NULL) :
3691 			    pfil_mbuf_in(V_inet6_pfil_head, mp, ifp, NULL);
3692 			if (rv != PFIL_PASS)
3693 				break;
3694 		}
3695 
3696 		if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3697 		    pfil_mbuf_in(V_inet6_pfil_head, mp, bifp, NULL)) !=
3698 		    PFIL_PASS)
3699 			break;
3700 		break;
3701 #endif
3702 	}
3703 
3704 	switch (rv) {
3705 	case PFIL_CONSUMED:
3706 		return (0);
3707 	case PFIL_DROPPED:
3708 		return (EACCES);
3709 	default:
3710 		break;
3711 	}
3712 
3713 	error = -1;
3714 
3715 	/*
3716 	 * Finally, put everything back the way it was and return
3717 	 */
3718 	if (snap) {
3719 		M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3720 		if (*mp == NULL)
3721 			return (error);
3722 		bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3723 	}
3724 
3725 	M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3726 	if (*mp == NULL)
3727 		return (error);
3728 	bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3729 
3730 	return (0);
3731 
3732 bad:
3733 	m_freem(*mp);
3734 	*mp = NULL;
3735 	return (error);
3736 }
3737 
3738 #ifdef INET
3739 /*
3740  * Perform basic checks on header size since
3741  * pfil assumes ip_input has already processed
3742  * it for it.  Cut-and-pasted from ip_input.c.
3743  * Given how simple the IPv6 version is,
3744  * does the IPv4 version really need to be
3745  * this complicated?
3746  *
3747  * XXX Should we update ipstat here, or not?
3748  * XXX Right now we update ipstat but not
3749  * XXX csum_counter.
3750  */
3751 static int
3752 bridge_ip_checkbasic(struct mbuf **mp)
3753 {
3754 	struct mbuf *m = *mp;
3755 	struct ip *ip;
3756 	int len, hlen;
3757 	u_short sum;
3758 
3759 	if (*mp == NULL)
3760 		return (-1);
3761 
3762 	if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3763 		if ((m = m_copyup(m, sizeof(struct ip),
3764 			(max_linkhdr + 3) & ~3)) == NULL) {
3765 			/* XXXJRT new stat, please */
3766 			KMOD_IPSTAT_INC(ips_toosmall);
3767 			goto bad;
3768 		}
3769 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
3770 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3771 			KMOD_IPSTAT_INC(ips_toosmall);
3772 			goto bad;
3773 		}
3774 	}
3775 	ip = mtod(m, struct ip *);
3776 	if (ip == NULL) goto bad;
3777 
3778 	if (ip->ip_v != IPVERSION) {
3779 		KMOD_IPSTAT_INC(ips_badvers);
3780 		goto bad;
3781 	}
3782 	hlen = ip->ip_hl << 2;
3783 	if (hlen < sizeof(struct ip)) { /* minimum header length */
3784 		KMOD_IPSTAT_INC(ips_badhlen);
3785 		goto bad;
3786 	}
3787 	if (hlen > m->m_len) {
3788 		if ((m = m_pullup(m, hlen)) == NULL) {
3789 			KMOD_IPSTAT_INC(ips_badhlen);
3790 			goto bad;
3791 		}
3792 		ip = mtod(m, struct ip *);
3793 		if (ip == NULL) goto bad;
3794 	}
3795 
3796 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3797 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3798 	} else {
3799 		if (hlen == sizeof(struct ip)) {
3800 			sum = in_cksum_hdr(ip);
3801 		} else {
3802 			sum = in_cksum(m, hlen);
3803 		}
3804 	}
3805 	if (sum) {
3806 		KMOD_IPSTAT_INC(ips_badsum);
3807 		goto bad;
3808 	}
3809 
3810 	/* Retrieve the packet length. */
3811 	len = ntohs(ip->ip_len);
3812 
3813 	/*
3814 	 * Check for additional length bogosity
3815 	 */
3816 	if (len < hlen) {
3817 		KMOD_IPSTAT_INC(ips_badlen);
3818 		goto bad;
3819 	}
3820 
3821 	/*
3822 	 * Check that the amount of data in the buffers
3823 	 * is as at least much as the IP header would have us expect.
3824 	 * Drop packet if shorter than we expect.
3825 	 */
3826 	if (m->m_pkthdr.len < len) {
3827 		KMOD_IPSTAT_INC(ips_tooshort);
3828 		goto bad;
3829 	}
3830 
3831 	/* Checks out, proceed */
3832 	*mp = m;
3833 	return (0);
3834 
3835 bad:
3836 	*mp = m;
3837 	return (-1);
3838 }
3839 #endif /* INET */
3840 
3841 #ifdef INET6
3842 /*
3843  * Same as above, but for IPv6.
3844  * Cut-and-pasted from ip6_input.c.
3845  * XXX Should we update ip6stat, or not?
3846  */
3847 static int
3848 bridge_ip6_checkbasic(struct mbuf **mp)
3849 {
3850 	struct mbuf *m = *mp;
3851 	struct ip6_hdr *ip6;
3852 
3853 	/*
3854 	 * If the IPv6 header is not aligned, slurp it up into a new
3855 	 * mbuf with space for link headers, in the event we forward
3856 	 * it.  Otherwise, if it is aligned, make sure the entire base
3857 	 * IPv6 header is in the first mbuf of the chain.
3858 	 */
3859 	if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3860 		struct ifnet *inifp = m->m_pkthdr.rcvif;
3861 		if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3862 			    (max_linkhdr + 3) & ~3)) == NULL) {
3863 			/* XXXJRT new stat, please */
3864 			IP6STAT_INC(ip6s_toosmall);
3865 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3866 			goto bad;
3867 		}
3868 	} else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3869 		struct ifnet *inifp = m->m_pkthdr.rcvif;
3870 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3871 			IP6STAT_INC(ip6s_toosmall);
3872 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3873 			goto bad;
3874 		}
3875 	}
3876 
3877 	ip6 = mtod(m, struct ip6_hdr *);
3878 
3879 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3880 		IP6STAT_INC(ip6s_badvers);
3881 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3882 		goto bad;
3883 	}
3884 
3885 	/* Checks out, proceed */
3886 	*mp = m;
3887 	return (0);
3888 
3889 bad:
3890 	*mp = m;
3891 	return (-1);
3892 }
3893 #endif /* INET6 */
3894 
3895 #ifdef INET
3896 /*
3897  * bridge_fragment:
3898  *
3899  *	Fragment mbuf chain in multiple packets and prepend ethernet header.
3900  */
3901 static int
3902 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3903     int snap, struct llc *llc)
3904 {
3905 	struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3906 	struct ip *ip;
3907 	int error = -1;
3908 
3909 	if (m->m_len < sizeof(struct ip) &&
3910 	    (m = m_pullup(m, sizeof(struct ip))) == NULL)
3911 		goto dropit;
3912 	ip = mtod(m, struct ip *);
3913 
3914 	m->m_pkthdr.csum_flags |= CSUM_IP;
3915 	error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3916 	if (error)
3917 		goto dropit;
3918 
3919 	/*
3920 	 * Walk the chain and re-add the Ethernet header for
3921 	 * each mbuf packet.
3922 	 */
3923 	for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3924 		nextpkt = mcur->m_nextpkt;
3925 		mcur->m_nextpkt = NULL;
3926 		if (snap) {
3927 			M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3928 			if (mcur == NULL) {
3929 				error = ENOBUFS;
3930 				if (mprev != NULL)
3931 					mprev->m_nextpkt = nextpkt;
3932 				goto dropit;
3933 			}
3934 			bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3935 		}
3936 
3937 		M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3938 		if (mcur == NULL) {
3939 			error = ENOBUFS;
3940 			if (mprev != NULL)
3941 				mprev->m_nextpkt = nextpkt;
3942 			goto dropit;
3943 		}
3944 		bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
3945 
3946 		/*
3947 		 * The previous two M_PREPEND could have inserted one or two
3948 		 * mbufs in front so we have to update the previous packet's
3949 		 * m_nextpkt.
3950 		 */
3951 		mcur->m_nextpkt = nextpkt;
3952 		if (mprev != NULL)
3953 			mprev->m_nextpkt = mcur;
3954 		else {
3955 			/* The first mbuf in the original chain needs to be
3956 			 * updated. */
3957 			*mp = mcur;
3958 		}
3959 		mprev = mcur;
3960 	}
3961 
3962 	KMOD_IPSTAT_INC(ips_fragmented);
3963 	return (error);
3964 
3965 dropit:
3966 	for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3967 		m = mcur->m_nextpkt;
3968 		m_freem(mcur);
3969 	}
3970 	return (error);
3971 }
3972 #endif /* INET */
3973 
3974 static void
3975 bridge_linkstate(struct ifnet *ifp)
3976 {
3977 	struct bridge_softc *sc = NULL;
3978 	struct bridge_iflist *bif;
3979 	struct epoch_tracker et;
3980 
3981 	NET_EPOCH_ENTER(et);
3982 
3983 	bif = ifp->if_bridge;
3984 	if (bif)
3985 		sc = bif->bif_sc;
3986 
3987 	if (sc != NULL) {
3988 		bridge_linkcheck(sc);
3989 		bstp_linkstate(&bif->bif_stp);
3990 	}
3991 
3992 	NET_EPOCH_EXIT(et);
3993 }
3994 
3995 static void
3996 bridge_linkcheck(struct bridge_softc *sc)
3997 {
3998 	struct bridge_iflist *bif;
3999 	int new_link, hasls;
4000 
4001 	BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
4002 
4003 	new_link = LINK_STATE_DOWN;
4004 	hasls = 0;
4005 	/* Our link is considered up if at least one of our ports is active */
4006 	CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
4007 		if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
4008 			hasls++;
4009 		if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
4010 			new_link = LINK_STATE_UP;
4011 			break;
4012 		}
4013 	}
4014 	if (!CK_LIST_EMPTY(&sc->sc_iflist) && !hasls) {
4015 		/* If no interfaces support link-state then we default to up */
4016 		new_link = LINK_STATE_UP;
4017 	}
4018 	if_link_state_change(sc->sc_ifp, new_link);
4019 }
4020